UpdateNotes.txt

    Update notes for MadGraph5_aMC@NLO (in reverse time order)

ANNOUNCEMENT: 
   The 2.9.X LONG TERM STABLE version is now at end of life for bug fixing/support since December 2025.
   A new LTS (based on the current 3.5.X) is starting now and will act as the stable release for the coming years.

0.3.0 (XX/XX/XX):
   OM: madevent MLM (ickkw=1) with use_syst: the <pdfrwt> entry of the incoming partons
       used the flavour class of the grouped subprocess instead of the event's flavour
       (pdg 1 for s s~, 2 for c c~), so the systematics (PDF and mu_F variations) of every
       s/c-initiated event were computed with the d/u PDF (up to 9% for mu_F=2 on p p > z).
   OM: madevent LO vector mode (vector_size>1): the good-helicity scan saw a single
       beam orientation per warp, so up to half of p p > z was dropped. With helicity
       recycling, the madevent_forhel scan now runs event by event. Without it
       (hel_recycling = False), the scan inside SMATRIX now runs its MAXTRIES events
       for each beam orientation and gets the orientation of each event. Also fixed a
       latent bug: an event in IMODE 5 no longer skips the rest of the vector (the
       vector code currently always uses IMODE 0).
   RF: Add real-diagram channels for fixed-order NLO, through the
       hidden option real_channels_x0 in the run_card. Useful for
       processes that have new structures opening beyond LO, such as
       pp>w+bb [QCD] (W radiating from a final state quark line) or
       pp>w+h [QED] (photon induced t-channel contribution). A typical
       value would be a few percent or so (e.g., 0.03). See also
       Sec.6.3 of 0908.4272 [hep-ph].
   OM: real_channels_x0 is now set by default, at output, for the processes
       whose real-emission diagrams have propagators that the Born channels
       do not follow (0.05, e.g. pp>w+h [QED], pp>w+bb~ [QCD], pp>w+w- [QED];
       -1 otherwise, e.g. pp>jj, pp>w+j, pp>tt~j, pp>tt~h [QCD]). The
       decision is the same for a resonance and its decay products
       (pp>w+j and pp>e+vej). The value is written in the run_card, and
       ignored for MC@NLO/LO/merged runs.
   OM: The default PDF follows the process when a photon is needed (issue #311):
       a photon in the initial state of a proton beam (LO, NLO and the mg7
       [beam] pdf) or NLO EW corrections ([QED]) select
       NNPDF40_nlo_as_01180_qed (lhaid 335900), at NLO also for a model with a
       massive b (NNPDF4.0 has no 4-flavour QED set). The NNPDF4.0 defaults
       have no photon. A PDF set by the user is kept.
   OM: new command 'install cuttools_iregi' recompiling CutTools and IREGI (vendor/)
       with the current fortran compiler. The error raised when CutTools/IREGI were
       compiled with another compiler (e.g. after a gfortran update) now points to it.
   OM: compute_widths (and the 'auto' widths of a param_card) integrates the
       N-body decays (N>2) with madspace instead of MadEvent: 6 s instead of
       26 s for the Higgs of heft up to 3 bodies, with a more precise result.
       'compute_widths ... --engine=madevent' (or 'DECAY 6 auto
       --engine=madevent') selects the former MadEvent integration. Falls
       back to MadEvent, with a warning, if madspace cannot do it. The precompiled
       standard headers (new 'make PCH=1' of madmatrix, clang only) halve the
       build of many small processes: used by mg7 runs of 4 subprocess
       directories or more.
   OM: compute_widths --nlo without a dedicated code for the model (SMWidth, the
       SM) now runs the two-body decays with a colour charge at NLO with MC@NLO
       (aMC@NLO fixed order, loop_sm for sm); the decays into 3 bodies or more
       stay at LO. 'DECAY 6 auto --nlo --body_decay=3' takes the options of
       compute_widths in the param_card ('auto@NLO' stays 'auto --nlo'), also
       with 'set decay 6 auto --nlo --body_decay=3'. A width on auto that is
       not computed in the same call is now kept in the card (it used to be
       dropped when another width was computed). '--nlo=smwidth' and
       '--nlo=mcatnlo' choose the code (SMWidth, QCD and EW, or MC@NLO, QCD);
       the QCD widths of the two agree to 0.05% for t, W and Z.
   OM: compute_widths --nlo --corrections=QED|QCD+QED runs the two-body decays at
       NLO electroweak (or QCD and EW) with MC@NLO ([QED] / [QCD QED] of aMC@NLO,
       loop_qcd_qed_sm), for every two-body decay and not only the coloured ones.
       t, W and Z agree with SMWidth to 0.1% (EW) and 0.13% (QCD+EW, the top
       without the finite width of the W); the b is run apart from d and s since
       its isospin partner is the top. Bug fix: the NLO runs used the names of the
       UFO model for the particles (m-, tt+) and failed for leptons: they use the
       names of MG (mu-, ta+).
   OM: Bug fix: 'compute_widths --nlo' with SMWidth gave zero for every width
       when the path of the param_card was longer than 100 characters (smwidth
       cannot read it): it now works on copies with short names, and reports
       an error when smwidth gives no width.
   OM: Bug fix: the alpha_s of a madspace decay (1 > N, also 'output mg7' of a
       decay) was alpha_s(m_Z) at any mass. It is now the 2-loop running up to
       the decaying mass, as in MadEvent: e.g. 4.7% lower alpha_s(125 GeV) and
       9.7% lower width for h > g g in heft.
   OM: Bug fix: 'check_sa.exe perf' of 'output standalone' ignored the alpha_s
       of the param card and always used 0.118 (the 'matrix' mode was right).
       With --momenta/-e and --dump-me, the matrix elements at user points are
       now evaluated with the card alpha_s.
   OM: Bug fix: in 'output standalone_fortran' (without --prefix), a plain
       'make' in SubProcesses built the python module of the first P directory
       only. It now builds all of them, like 'make all'.
   OM: multi_run:the merge of the runs (merge.pl) is now done in python, 5 to 10
       times faster. Fix the normalisation for event_norm=sum (the merged sample was
       N_run times too large) and the <rwgt> weights (systematics/reweighting), which
       were not rescaled with the central weight. The relative weights of the events
       (bias, MadSpin, ...) are kept instead of being set to the same unit weight.
   OM: 'import model' now says when the model keeps a mass for u, d, s, e or mu:
       such a mass is negligible at collider energies but brings in Yukawa
       vertices and helicity configurations in every process, and
       'customize_model' removes it. Silent for a massive b or tau, which is
       the 4F scheme and a deliberate choice.
   OM: Bug fix: mg7 runs on AMD GPUs (device = ["hip"]) aborted at the start of
       the survey with "GPU error: an illegal memory access was encountered".
       The GPU matrix element rounds the batch up to a multiple of 256 events
       and computes all of them, but only the first 'count' events of the
       padded batch received inputs. The flavour index of the padding events
       was whatever the fresh allocation held, and cFlavors[iflavor] then read
       far out of bounds. The padding events are now copies of the first event.
       CUDA has the same code; it did not crash there only because the fresh
       memory happened to hold harmless values.
   OM: mg7 (madspace) now supports the run_card "me_frame", with the same meaning
       as in madevent: it names the external particles whose momentum sum is put
       at rest, and the matrix element is evaluated on momenta boosted into that
       frame. Only the matrix element sees them -- the cuts, the scales, the
       momentum fractions and the written event all keep the momenta as
       generated. It only changes the result for a matrix element that is not
       Lorentz invariant, i.e. a polarised one, and it is then the frame the
       polarisations are defined in; the default [] applies no boost, so every
       other run is unaffected and pays nothing. [1, 2] asks for the partonic
       centre of mass, which is the frame madevent evaluates in, and [1] is its
       equivalent for a 1 -> n decay. madspace generates its momenta in the lab
       frame, so unlike madevent it reaches the requested frame in two steps,
       through the rest frame of the incoming system: a single boost from the
       lab would leave a Wigner rotation behind and pick different polarisation
       axes than madevent.
   OM: BUG FIX (reweight): for a sample generated with a non-default me_frame,
       the matrix elements were still evaluated in the partonic centre of mass.
       The frame was read from frame_id, which a run_card read back from the LHE
       banner never sets (it always said 6, whatever me_frame was), and which an
       mg7 banner has none of -- there me_frame = [] means no boost at all, the
       matrix element having seen the lab-frame momenta. It is now read from
       me_frame, LO, NLO and mg7 alike. Only polarised samples were affected.
       RunCardMG7.from_LO now also carries me_frame over, instead of reporting
       it as unsupported. MadSpin likewise takes the frame of an mg7 sample from
       its run.me_frame, the default [] meaning the lab frame (no boost) rather
       than the partonic centre of mass every other sample defaults to.
   OM: BUG FIX (mg7): the initial-state mirror paired each event with a matrix
       element evaluated on the momenta *before* the mirror was applied. madspace
       covers the beam-swapped copy of an asymmetric initial state by rotating
       the event by pi about x (py, pz -> -py, -pz), but only the written event
       was rotated; the weight kept the momenta as generated. The two agree only
       because |M|^2 is invariant under that rotation -- which a polarised matrix
       element is not once it is evaluated in a frame that holds the polarised
       particle at rest, because HELAS then quantises the spin along the frame z
       axis, the mirror flips it, and the + and - states swap. p p > z{+} j and
       p p > z{-} j evaluated in the Z rest frame came out 4485 +- 39 pb against
       6060 +- 54 pb, 23 sigma apart, where the symmetry of a pp collider
       requires them to be equal; they now agree within 1 sigma. The sum over the
       three polarisations was already right, which is why no unpolarised result
       is affected (checked: p p > z j unchanged at 12420 +- 69 pb). The matrix
       element is now evaluated on the momenta the event is written with, which
       is both what madevent does and what the systematics reweighting here
       already assumed, since it reads them back out of the event buffer.
   OM: 'customize_model' now works for every UFO model: generic options (flavour
       scheme, massive leptons, diagonal CKM), commands to fix a parameter or a whole
       block ('set decay all 0'), to change a formula, to choose the values written in
       param_NAME.dat ('set default'), and --explain to say what each choice removes.
   OM: New command 'explain_restriction': report what a restriction card removes.
   OM: Bug fix: a parallel build of a generated Source/ could fail and still be
       reported as a success. Every library rule read 'cd DIR; make; cd ..', so
       the recipe returned the exit status of the final 'cd' -- always 0 -- and a
       failed sub-make went unnoticed until a subprocess directory later died
       with "No rule to make target '../../lib/libdhelas.a', needed by 'gensym'".
       They now use 'cd DIR && make'.
       The underlying failure was a 'make -j' race on Fortran module files: the
       Majorana-flip DHELAS routines and PDF/pdfwrap_lhapdf.f 'use model_object'
       (from MODEL/flavor_couplings.f), the other MODEL sources use it too, and
       the ALOHA routines 'use aloha_object' (from DHELAS/aloha_functions.f).
       Order-only prerequisites now enforce those three orderings.
   OM: 'set madanalysis_path'/'set td_path' no longer raise: the options are gone
       with MadAnalysis4, so a value that would only disable them is ignored
       silently and any other value warns once and is dropped. Stale entries in an
       old configuration file are dropped the same way. This keeps old command
       files and old process directories working.
   OM: Removed MadAnalysis4 support. The 'install MadAnalysis4' command, the
       'madanalysis_path'/'td_path' configuration options, the 'plot' command of
       both madevent and aMC@NLO, the plot_card.dat and the topdrawer-based
       plotting of the parton/Pythia6/PGS/Delphes outputs are all gone. The
       topdrawer ('td') binary it relied on is no longer downloadable, so the
       install had been broken for a while. Use MadAnalysis5 or Rivet instead.
       The Pythia8 merging (DJR/pt) plots are unaffected: they go through HwU.
   OM: Bug fix: OrganizeModelExpression.track_dependant is a class attribute that
       analyze_parameters() was extending in place, so the tracked variables of one
       UFO model leaked into every model imported afterwards in the same process.
       Importing a Gmu-scheme model (no external aEWM1) added 'Gf' to the class
       list for good, and a later import of e.g. loop_sm then grouped mdl_MW under
       ('aEWM1','Gf') instead of ('aEWM1',), reordering mp_intparam_definition.inc.
       Each instance now works on its own copy.


3.8.2 (XX/XX/XX):
    OM: LO grouped output (default): the renormalisation/factorisation scales and the MLM xqcut
        veto were computed on the momenta before the symmetry permutation of the channel (#453).
        Channels with such permutations were biased (e.g. -3.5% for u q > z u q, also with MLM).
        Fixed both for the scalar and the vector (vector_size>1, cudacpp) mode.
    OM: LO vector mode (vector_size>1): fix the reweighting of MLM matching (the central factorisation
        scale of the last event of the batch was used for all events), the config/mirror used per warp,
        and a crash (memory corruption) when vector_size was changed in the run_card after the output.
    OM: LO vector mode with nb_warp>1: when a P directory groups several subprocesses, each warp
        selected its own subprocess but only the last one was evaluated for all warps (wrong weights,
        'Clustering failed' crashes). The subprocess is now chosen once per vector (from the PDF weights
        averaged over the warps) and the configuration/beam orientation per warp.
    OM: MadSpin (spinmode=PA): new option 'auto_zero_width_PA' (default True) setting the width
        of the particles decayed by MadSpin to zero in the matrix element of the production density
        matrix, and each one in the matrix element of its own decay density matrix only (Breit-Wigner
        sampling, reshuffling, propagator factor and branching ratios keep the param_card widths).
    OM: output standalone: new option --zerowidth_decay=True|False|<pdg>,<pdg>,... writing propagators
        with a zero width (param_card untouched): in a 2>N process those of the final-state particles,
        in a 1>N process (decay) those of the decaying particle only; a pdg list restricts to those.
    VD+OM: Density mode of the reweighting (since 3.8.0): the particles of the density matrix are now
        ordered as given in particle_in_density_matrix (they followed the order of the process), also
        when they are ranked with an observable. That ranking could be wrong, or drop a particle, when
        other particles were in between. A pdg code (or a rank) absent from the process is now an error.
    OM: Density mode of the reweighting (since 3.8.0): boost_choice used the wrong particles when the
        event contains intermediate particles, e.g. the t+W frame instead of the b+t~ one for
        g g > t t~, t > b w+ with boost_choice [5, -6].
    VD: Quantum Information library: new observable Get_Steering (two-qubit steering for 2 or 3
        measurements per site, 1510.08030).

3.8.1 (XX/XX/XX):
    OM: LO: the final unweighting no longer truncates the events above the maximum
        weight; they keep the weight w/w_max in units of the common weight
        (part of the under-population of the tails).
    OM: LO: fix dsqrt_shatmax: the phase-space grids were set in the wrong units, giving a zero
        cross-section for 2>1 processes (e.g. p p > z) and a crash (Bjorken x > 1) when
        dsqrt_shatmax was above the collider energy. dsqrt_shatmax <= 0 now means no cut
        (was a zero cross-section).
    ALL: Include all bug fixes from LTS 3.5.17 (see below)

3.8.0 (15/09/26):
    OM+SA+VD: MadSpin2 (2609.02794)
        - Arbitrary decay, support for loop-induced processes, interference, Pole Approximation mode, polarization weights, ...

    FM+OM+VD: Quantum Information mode (2510.17730)
        - Computation of density matrix for tree level and loop-induced
        - Reweighting mode to get the density matrix over a sample of events
        - Quantum Information python library for easy analysis (including many QI variables)

    HS+LM+LS: P-Wave Support for Quarkonium/Leptonium (2607.26739)

    Bug Fixes:
        OM: Bug fix custom files for cuts/scale/... were broken for NLO computation
            since 3.6.0
        OM: Bug fix (since 3.6.0) for rapidity cut in asymmetric collision 
            (ebeam1 != ebeam2)
        BF+JT: Fix two independent loop-wavefunction bugs in 
            check_majorana_and_flip_flow for models with sextet/Majorana
        OM: MadSpin onshell_v1: wrong identical-particle factor between decay-chain
            and production ME
        OM: MadSpin v1: two beam-polarisation fixes (wrong polbeam convention, 
            reweighting applied after the helicity sum), both since 3.6.3
        OM: Beam polarisation applied twice when a beam helicity is fixed in the process
        OM: Long run_card lines (e.g. systematics_arguments) were truncated 
            in the LHE banner
        OM: Make Source/make_opts writes atomic to avoid corrupted files


3.7.3 (01/09/26):
    OM: MadSpin: the run_card me_frame and polbeam1/polbeam2 are now honoured by the
        density-based spinmodes (onshell/PA/madspin/full) and no longer only by the
        legacy madspin_v1/onshell_v1 paths.
        - beampol reweights the initial-state helicity sum inside the density matrix,
          through a new PY_SET_BEAMPOL entry point of the matrix-element library;
        - me_frame boosts the momenta of the production and of every decay into the
          selected frame before the matrix element sees them, which is what defines
          the axis those helicities are quantised along. The boost is skipped for
          unpolarised beams, so unpolarised runs are unchanged bit-for-bit. Note
          that -- in the density modes as in the v1 path -- the frame turns out not
          to change any observable: the initial-state legs are massless, so their
          helicity labels and each fixed-helicity |M|^2 with them are boost
          invariant, and so is the reweighted sum. What the boost does change is the
          basis the individual density-matrix elements are written in, which is why
          the production and every decay have to be boosted with the same momentum.
        - "set frame_id"/"set beampol" in the MadSpin card now override the run_card
          values instead of being silently discarded.
    OM: Two MadSpin bug fixes in the beam-polarisation support of the legacy
        madspin_v1/onshell_v1 paths (both since it was added in 3.6.3):
        - the run_card polbeam1/polbeam2 were converted to the [0,1] left-handed-
          fraction convention of the EVA PDF and then fed to a matrix-element
          reweighting that expects madevent's /to_polarization/ convention. As a
          result, an *unpolarised* run (polbeam=0 -> beampol=0.5) applied a spurious
          factor 0.5 to one initial-state helicity and 1.5 to the other, while a
          fully polarised beam (polbeam=-100 -> beampol=1) applied nothing at all,
          and the sign of the polarisation was flipped. The mapping is now
          sign(1+|polbeam|/100, polbeam), as in setrun.f;
        - in matrix_standalone_msF_v4.inc the reweighting was applied after the
          helicity sum rather than before it, so only HELAMP was polarised and the
          full matrix element -- the numerator of the MadSpin decay weight, and the
          only place the polarisation can reach the decay angles -- stayed
          unpolarised. The decay distribution of the v1 path was therefore
          insensitive to any partial beam polarisation (100% happened to work by
          accident, through the GOODHEL filter zeroing the disfavoured helicities).
    OM: BUG FIX (MadSpin): when two or more final-state particles with the same PDG code were decayed
        via *different* decay lines -- e.g. "p p > z z" with "decay z > e+ e-" and "decay z > u u~" --
        the two decay branches were paired with the two mothers in reverse order. The spin-correlated
        weight of a given mother was therefore evaluated with one decay branch while the decay products
        of the other branch were attached to it. Decay angular distributions were consequently wrong
        (in p p > z z the lepton/quark helicity angle is off by ~45 sigma on a 200k event sample), while
        the cross-section, the branching ratios, the decay-plane angle and the spin-correlation
        observable <cos(theta_1) cos(theta_2)> were left exactly invariant -- which is why this stayed
        unnoticed. Processes where each decayed particle carries a distinct PDG code (t t~, w+ w-, ...)
        were never affected.
    OM: BUG FIX (polarisation): LO cross-sections computed with a run_card "me_frame" that selects a
        single particle -- e.g. "me_frame = 3" to work in the Z rest frame, which is the standard way
        to ask for the polarisation of that Z -- were wrong for a fraction of the events.
    OM: Allow to have +,- polarization restriction for onshell massive boson propagator 
    RF: Fix an issue in the collection of events at NLO when using too many cores (introduced in 3.7.2)
    RF: Allow to have HWU histograms render with matplotlib instead of gnuplot
    OM: improved fix for a missing LHAPDF argument in the info file


3.7.2 (07/07/26):
    MZ: negative seeds are now allowed and are strictly pinned (so no automatic reset to 0 for the following run).
        negative seeds are identical to positive ones (both produce exactly the same sample)
    RF: Fixed the writing of LHE files in fixed-order computations that was broken since the previous release (3.7.1).
    OM: Ensure that goldstones are merged with the SM coupling for FD gauge.
        Note that FD needs models that are built such that the renormalization does not induce a shift of the SM values
    OM: Allow the possibility to run Delphes in multicore. Compatible with the hepmc mode of pythia8 or autoremove
        to avoid the costly merging of hepmc files.
    OM: introduce two options nb_core_pythia8 nb_core_delphes
    OM: Fix systematics.py for pdf variation for pdf set different than the original (thanks to Z. Marschall)
    OM: (Try) to automatically update a pdf set if the pdf set has missing keys (like AlphaS_NumFlavors)
    Team: include all bug fixes from 3.5.16 (see below)

3.7.1 (29/04/26):
    OM: Change the handling of seed at NLO, after a run with a given seed, the seed is reset in the run_card 
        to 0 (automatic) meaning that the next run in the directory will be with a different seed. 
        Note that this was the behavior at LO already.
    OM: Allow easier syntax in the run_card. 
        You can now use "set lpp1 e-" instead of "set lpp1 3"
        You can now use "set fixed_scale mz" instead of "set fixed_scale 91.1880"
        You can now use "set dynamical_scale_choice ht/2" instead of "set dynamical_scale_choice 3"
    OM: Change the implementation of the "$" syntax such that the implementation moves at the amplitude level
        This means that the syntax is now also supported at standalone level (bwcutoff hardcoded to 15)
        This has no impact if sde_strategy=1 but avoids issue with sde_strategy=2.
        We still recommend sde_strategy=1 when using the $ syntax since you can have spurious writing of resonances
        in the lhef with sde_strategy=2.
    RF: Refactored the code that collects all the events from the various channels
        into a single event file at the end of an MC@NLO run.
    OM: Remove the dependency on the six package
    MZ+LJ: Bug fix in ew_sudakov (one in the color matrix and one in the value for alphas)
    OM: Fix a couple of issues with the linking of the heptools library (especially on ubuntu)
    RR: Extended propagator support for polarised bosons (https://arxiv.org/abs/2512.10015)
        See https://github.com/mg5amcnlo/mg5amcnlo/pull/215 for simple instructions
    OM: Cleaner version of how MLM selects the color for events
    Team: include all bug fixes from 3.5.14 (see below)
        Note slightly different handling of default parameter when a parameter is not within the run_card

3.7.0 (05/01/26):
    RR: When carrying out parameter scans in MadGraph/MadEvent or MG5aMC@NLO with systematics (use_syst True or reweight_scale/pdf True), then the uncertainty information is also recorded in the scan summary.
    RR: Enhance the equivalent vector approximation (EVA) implementation by adding full leading-power and next-to-leading-power accuracy modes, plus an optional xcut kinematic restriction, refactor core routines in both Fortran and Python, and introduce new (hidden by default) run_card parameters 'eva_xcut' to enable or disable the kinematic restriction x>MV/Ebeam in EVA.
    LS: Support for NLO calculations in ultra-peripheral collision (UPC) processes (see arXiv:2504.10104)
       Details: Introduce tagged initial-state photon (with symbol !a!) to denote coherent incoming photons in UPCs
       → Prevents generation of ISR diagrams involving initial-state fermion splitting
       → When QED corrections are requested, adopt the hybrid renormalization scheme (same as for tagged final-state photons)
       → For Born processes containing jets j, replace coherent photons !a! with the multiparticle aUPC, which includes all jet fermions plus the photon, ensuring all underlying Born diagrams required for subtraction are generated
    SR: adding support for multi-gpu
    

3.6.7 (05/01/26):
     ALL: include bug fixes for the 2.9.x LTS (latest bug fix for that LTS: 2.9.27) and from the new LTS 3.5.12
     OM: improvement of the help command (the manual) for the set command.
         Now any "set" command has its own help via "help set <name>"
     GabrielMajeri: Add support for pigz (parallel version of gzip) speeding up the code 
     jakobnov: Adding support for checkpointing at NLO (with DMTCP), see "help set checkpointing" for details.
     MZ: Fix another issue with the color matrix (introduced in 3.6.2) for computations with an interference term
     OM: fix a crash for loop-induced processes (related to the dilog function defined twice --introduced in 3.6.6--)

3.6.6 (30/10/25):
     OM: Fix a serious bug in the lhe writing of events where some intermediate particles were wrongly written in the file.
     OM: change in default restriction for coupling order for BSM (closer to 2.9.x behaviour)
     OM: new plugin feature: possibility to customize the entry in the scan summary file
     OM: Fix the support of some electron/muon PDFs at LO
     OM: User using git will now have the git version/tag printed in the banner of MG5aMC instead of a warning
    
3.6.5 (17/10/25):
     MZ: Ensure that the code is crashing at NLO when the analysis contains a syntax error
     OM: add a "compile" command at LO
     OM: Fix a bug in cudacpp where the color of the events was wrongly selected when writing the events into the lhef file

3.6.4 (13/9/25):
     RF: Fixed critical bug in FxFx merging, introduced in version 3.3.1. This resulted in too many negatively weighted events,
         and too many high-pT jets. Particularly relevant for merging with up to 2 or 3 jets at NLO. Thanks to Lenart Jerala
	 for pointing this out and testing the fixes.

    ALL: Merge with 3.5.10/2.9.25 bug fix (see below)
         Include a change in the polarization reported for intermediate particles
        (9 reported now instead of 0)

3.6.3 (12/6/25):
     OM: Fix conversion model for the FD gauge
     OM: Include all bug fixes from 3.5.9 and 2.9.24
     OM: Fixing a bug that EW was by default on NLO+PS mode and not fixed order

3.6.2 (19/3/25):
      RF: Implemented flavour_bias parameter in the NLO run-card to allow for biased event generation
          based on the flavours of the external particles, as used in 2403.14419.
      RF: Use a grid for the running of alpha_S (NLO only and only when not using LHAPDF). 
      Andy Buckley: change jpg output by png output for better quality (with OM)
      RF: Updated histograms.py to prevent warnings/errors with latest gnuplot version
      OM: Update Color computation (that part of the computation is close to two times faster and 
          much faster to compile)

3.6.1 (29/11/24):
      OM+MZ: Fix a bug for lhapdf at NLO where the code was incorrectly changed like the LO version (and was crashing at compilation)
      OM: Fixing various compilation issues at LO
      ALL: Include fixes from LTS 3.5.7 and 2.9.22
           - including some rare but serious bug in scan
           - including support for python 3.13 (but for f2py related feature)

3.6.0 (30/09/24):
      PT+RF: fixed the Pythia8 script so that SCALUP in the LHE file is ignored only when using FxFx or MC@NLO-Delta
      OM: Allow the scan command within the run_card.
          Note that such command can NOT be used if the scan command is used in the param_card.
          so the following command is now working:
          set ebeam scan1:[6500,7000]
          Thanks to Kentarou Mawatari for pushing me to implement this.
      DP+TV+MZ: Implementation of EW Sudakov corrections. See 2110.03714 and 2309.00452
      AV+SR+SH+ZW+DM+OM: Implementation of the support for batch of events at LO.
          This will allow to support the cudacpp plugin that will offer
            - GPU support for LO processes
            - SIMD support on cpu
          Those modes need to install the cudacpp plugin (can be done via the command "install cudacpp")
          This feature also add OpenMP support for LO process (deactivated by default and which does not need the plugin).
      OM+JK+KM+KH+YJZ: implementation of the FD gauge via "set gauge FD" [2203.10440, 2405.01256]


3.5.17 (9/10/26):
  RF: Fixed a bug in FxFx merging: MG side used inclusive-kT measure, PY side used exclusive.  This mismatch resulted in an
      underrepresentation of events in DRjj<1.0, particularly visibile in boosted configurations. Thanks, Sihyun Jeon.
  OM: LO: setting mmnl without mmnlmax (default -1) silently removed the mmllmax upper cut
      for processes with exactly two leptons/neutrinos.
  OM: Fix beam polarization being applied twice when the helicity is fixed in the process definition
      (e.g. "ve{L} d > e- u" was twice too large with polbeam1=-100) [bug #2148181]
  OM: (NLO) Fix the truncation of long run_card lines (e.g. systematics_arguments) in the LHE banner [bug #2046152]
  OM: MadSpin: fix decay branches assigned in reverse order to identical particles with different decays
      (e.g. p p > z z with "decay z > e+ e-" and "decay z > u u~", also within a decay chain). Cross-section unchanged but decay
      angular distributions were wrong.
  OM: Reweighting: fix the crash when a non-default me_frame was used. The frame is now built as in madevent
      (from the partonic center of mass) and used for both the original and the new matrix-element.
  OM: Fix lhe_parser crash (or wrong output) when an event only matches the charge-conjugated process
  OM: Fix FourMomentum.boost when the boost vector is at rest (e.g. partonic frame of a lepton collider)
  OM: ALOHA: fix the sign of even power of the momentum for outgoing fermion in custom propagator
  OM: LO: the s-hat sampling starved the region below M-5*Gamma of every resonance not in a decay chain
      (e.g. p p > e+ ve j with mmnlmax=40 was 25% too low) and nested resonances (t > b w+, w+ > e+ ve)
      were counted twice (single top p p > e+ ve b j a few percent too low).
  OM: LO: fix the propagator denominators of the multi-channel weights of sde_strategy=2 (mass and mass
      squared were mixed and the width entered as Re(1/D^2) instead of 1/|D|^2). The weights did not follow
      the resonances, leading to slow integration and biased results (e.g. p p > e+ e- j 3.5% too low,
      single top 1% too low).
  RF: (NLO) Fix compilation of every NLO process with gfortran 16 (virtgranny_red not declared external
      in genps_fks.f)


3.5.16 (3/7/26):
  OM: fix a crash with systematics when the original pdf was not detected
  OM: fix some loop-induced compilation issues
  OM: avoid new Python syntax warnings

3.5.15 (17/4/26):
  OM: revert a change of 3.5.14 on the grid handling to be more secure on the change
  OM: Additional fix related to the matchbox template

3.5.14 (10/4/26):
  OM: Fix an issue in how sextets were written within the lhef output
  OM: Small change in the way default values are taken if some parameters are not present within the run_card
      Now the default value will be taken first from the run_card_default.dat (which is process specific) 
      rather than using a hardcoded default. This is required to correctly take sde_strategy=1 when
      a process uses the $ syntax (which is not compatible with sde_strategy=2). Note that 3.7.1 does change the
      "$" handling to avoid such a type of issue in a cleaner way.
  OM: Fixing various small issues related to linking to HEPTools
  OM: Change some API for the matchbox output (for the interference case)
  OM: Handle UFO models which are not python3.13 compatible (in the write_param_card.py script)
  OM: Change in how madgraph sets up the grid for shat, to better handle processes at high energy and with very soft cuts
      This also better handles the p p case since the grid is now aware of the implicit cut related to the CKKW merging scheme

3.5.13 (05/01/26):
  VD: fixing issue with custom functions that moving back to no custom functions was not correctly recompiling the code.
  OM: change f2py interface to make madspin/reweighting compatible both above and below python3.12
  OM: merge with the latest version of the 2.9.x LTS (2.9.27) 
      in particular fix a bug for event with discarded overweight (with RF)

3.5.12 (30/10/25):
  OM: Change in gensym to better determine which channels have an impossible BW configuration. This removes many spurious warnings.
  OM: include bug fixing from LTS (2.9.26) --see below --

3.5.11 (19/9/25):
  RF: Fixed critical bug in FxFx merging, introduced in version 3.3.1. This resulted in too many negatively weighted events,
      and too many high-pT jets. Particularly relevant for merging with up to 2 or 3 jets at NLO. Thanks to Lenart Jerala
	  for pointing this out and testing the fixes.

3.5.10 (13/9/25):
  All: propagate bug fixing from 2.9.x version (up to 2.9.25)

3.5.9 (4/6/25):
  OM: Fixed an issue in fixed_fact_scale when set to T where sometimes some beams were kept on False (in an asymmetric beam context)
  OM: Fixed a wrong scale assignment when beam#2 was dynamic but beam#1 was static (at LO)
  ALL: Include fix from 2.9.24 (including support to GCC15 + bug in madspin with GCC13) 
  OM: (thanks to SungBeom Cho) crash when running multi_run during the merging of multiple lhe files if
      MadSpin was activated (if the BR was different from one). If the BR=1, the merging was going through but
      the merged sample would contain both decayed and undecayed events, making that file pointless.

3.5.8:
  OM: Fix a quite serious bug introduced in 3.2.0:
      The scale choice can in some cases be assigned in an asymmetric way in the presence of mirror processes.
      This was leading to an asymmetric z -> -z production for LHC processes.
  OM: Fixing a gridpack issue that cross-section was wrongly reported in the comment of the banner
      Thanks to Joon-Bin-Lee and the CMS validation team for having identified those two bugs.
  OM: Include fix from LTS 2.9.23

3.5.7 (29/11/24)
  OM: Include fix from LTS 2.9.22

3.5.6 (26/9/24):
  OM: Include fix from LTS 2.9.21
  OM: fixing gridpack with large number of events (issue with the new strategy of combining events)

3.5.5 (17/06/24):
     OM: Fix a bug in the new strategy of combining events.
     OM: Allow slurm/condor cluster to specify the wall time of the jobs (cluster_walltime)
     OM: add new (hidden) entry in the run_card to force to keep all the log file (keep_log)
     OM: Fixing a matchbox issue (introduced in 2.9.0)
     OM: Dynamical scale choice for NLO can now be set via custom_fcts (was bugged before)
     OM: Include fixes from the (other) LTS 2.9.20

3.5.4 (05/04/23):
      MZ+RF: Bug fixes:
        1) a bug related to event-generation, when virtual corrections include more than one
        coupling combination (so NLO QCD to SM processes were **not** affected), in particular 
        when the coupling combinations have different sign. The contribution
        from the virtual amplitude was wrongly accounted for, resulting in unphysical distributions.
        Thanks to Haoquan Ren for reporting it.
        2) a bug related to NLO EW corrections at lepton colliders, for initial-state FKS configuration.
        Phase-space points for which the real-emission momentum generation was failing but the Born one
        was passing were thrown away altogether.
        Thanks to Riccardo Lubello for leading to the identification of the bug.
      OM: Change the way restriction models work, to be more aggressive in the data structure.
          This was triggered by the fact that the virtual was considered different if some unused Lorentz structure
          was left in the model/interaction. This allows to get back to the behaviour of the LTS version in terms 
          of number of directories for QCD processes and therefore gain disk area and speed-up the code.
      OM: Change in the unweighting strategy for processes with a large number of channels (>80).
          If you have deactivated the limit of number of open file (via the ulimit command) then the unweighting will be 
          done in a single step in one core (typically the most efficient strategy).
          If not possible, then the unweighting will be done by packets of 80 channels in a multi-core way.
          Then those packets will be merged together in a second step (single core).
          In this second case the algorithm has also been made more aggressive. 
          Speed-ups of a factor 20 have been observed in some extreme cases.
          Thanks to Mark Goodsell and Stephan Hageboeck, for pushing me on this and for your help.
      OM: Python3.12 status:
          - many (python) warnings are raised when running with python3.12. Work is in progress (and will be very long) to 
           remove all those warnings.
          - f2py (needed for reweighting/madspin) is quite impacted by the drop of distutils and typically fails to compile
            with the meson backend. numpy version 2.x (not yet released) will be the minimum for reweighting/madspin.
      OM: Allow (back) LHE output in fixed-order format (thanks Gauthier)
      OM: Change the internal representation of fixed_order card, that allow to use the set command to edit the card.
          like for the param_card/run_card/...
      OM: relax polarization restriction at NLO for massless particles.
      SJ+OM: fix an issue with not correctly saving where fastjet was installed when installing it via the MG5aMC install command
      SJ: Fixing rivet running, which was broken due to some libraries not being correctly linked
      OM: Include bug fix from 2.9.19

3.6.0:
      OM: Stop support for model in the legacy v4 format

3.5.3 (22/12/23)
      OM: Include Bug fix from 2.9.18
      OM: Update inner representation of the shower_card allowing to support any py8 parameter

3.5.2 (08/11/23)
      MZ: fixing issue for 2>1 processes at NLO accuracy (like p p > Z [QCD])
          For most compilers the process was just crashing, but some compilers were returning absurdly large cross-sections
      RF: update tests at NLO to better include phase-space cuts
      OM: Change restriction model to be less sensitive to numerical accuracy
      OM: bug fixing from LTS up to 2.9.17 (see below)
      PT: added options for matrix-element corrections in the shower card for PY8
      
3.5.1 (11/07/23)
      SH+OM: Update interface for contur
      OM: fix an issue with NLO reweighting introduced in 3.4.0 where loop scale was not correctly set anymore
      OM: Better determination of zero helicity for helicity recycling part of the code
      OM: Fix issue for powheg output
      OM: bug fixing from LTS up to 2.9.16 (see below) 
      
3.5.0 (12/05/23)
      VB+MC+SF+GS+MZ+XZ: NLO EW implemented for lepton-lepton collisions with ISR and beamstrahlung
	      See 2207.03265 and the FAQ https://answers.launchpad.net/mg5amcnlo/+faq/3324 for details. 
	      NNL-accurate parton distributions for lepton collisions are provided by eMELA 
	      https://github.com/gstagnit/eMELA/releases
      SF+MZ+PT+PY8author: Introduction of MCatNLO-Delta, a new matching procedure allowing to have less
          negative events generated (see 2002.12716)
      OM: MadSpin has now a syntax for handling semi-leptonic decay
          not supported in spinmode=onshell/none
      OM: possibility to install eMELA (install eMELA)
      HS: Adding support of UPC
      Bruno-Miguel-Oliveira: conform with the XDG Base Directory Specification for history/config file:
          https://github.com/mg5amcnlo/mg5amcnlo/pull/45
      OM: bug fixing from LTS up to 2.9.15 (see below)

3.4.2 (20/01/22)
      ALL: fixing bugs related to 2.9.13 (Long Term Stable: see below)
      OM: fix issue with hepmcremove (thanks to franaln)
      OM: fix some issue with addmodel
      OM: fix issue with the update of the param_card to set the used value for alphas
      OM: fix issue with wrong env variable for pythia8data (thanks pavel)

3.4.1 (01/09/22)
      MZ+CS: PineAPPL now supports user-defined bias_wgt_functions.
      RF: removed a write statement in the reweighting routine (to get scale and PDF uncertainties at NLO+PS runs) that cluttered the reweight_xsec_events.output log files.
      OM: Fixing an issue with lhapdf when trying to use a pdf set not yet installed within the database
      All: in sync with bug fixing in LTS (version 2.9.12)


3.4.0 (06/05/22)
      SJ: Allowing to use RIVET/CONTUR from madgraph. In the presence of a scan, such running can also be done
          at the end of the scan. (Contribution from Sihyun Jeon)
      OM: Allow EFT operators to run for some special UFO models (quite a restricted class of running is supported
          -- corresponding to EFT running --) 
      OM: change reweighting/spinmode=onshell (from madspin) to use the average of the matrix-element when it exists an order ambiguity in the matrix element like for the following process:
      p p > e+ ... z, z > e+ e-
      where permuting the momenta of the "e+" does not lead to the same
      matrix-element.
      CF: include notification when jobs are done in the unix notification center if executable notify-send is present.
          Contribution from Carlo Flore.
      All: Dropping the support for python2.7
      All: include bug fix from Long Term Stable version (2.9.10) see below

3.3.2(18/03/22)
      OM: (re)allow --loop_filter=True/... syntax for MC@NLO
      RF: Fixed a bug in the phase-space generation of MadSpin,
          relevant for processes with more than 1 t-channel.
      MZ: Restored functionalities for processes with external photons and 
        only QCD corrections
      all: include all bug fix from the LTS version (2.9.9)
           including important bug on MLM generation
	   
3.3.1 (04/12/21)
      RF: Fix for bug #1928364 related to the new FxFx clustering routines.
      OM: Fix for bug #1952050 systematics/reweighting run at NLO that was broken
          introduced in 3.3.0
      OM: Fix bug related to python 3.10
      OM: Fix bug when running from the process directory
      all: include all bug fix from the LTS version (2.9.7)
           including important bug on MC@NLO and on madspin


3.3.0 (12/11/21)
      DP+HS+IT+MZ: EW corrections can be computed for tagged photons
      OM+RR+AC: Implementation of EWA within MG5aMC (PDF for W/Z boson out of lepton beam)
      MZ+CS: A fix in amcatnlo_run_interface to prevent integer overflows when a
        very high precision is requested.
      MZ+CS: Fixes/improvements relevant to PineAPPL: 
            - the function pineappl_grid_optimize is called by pineappl_interface,
              to reduce the size of the grids. PineAPPL v0.5 or later is required
            - minor fix on the gluon pdg code in pineappl_interface.cc
            - the integration channels are treated the same way with or without PineAPPL
            (maxchannels is kept the same in both cases)
      RF: Fix for montecarlocounter in case the Born is not strictly positive
           (i.e., only interference contributions).
      RF+IT: Rewritten the clustering code for FxFx. This allows for
             separation of QCD and EW jets. Needs special JetMatching.h for
             Pythia8 (can be found in Template/NLO/MCatNLO/Scripts/JetMatching.h)
      OM: adding a new cut dsqrt_shat to set a minimum center of mass energy cut	   
      OM: allow the support of latest LUX PDF with lepton content of the proton
          Note that running PY8 is automatically forbidden in that case.
      ALL: include bug fixing from Long Term Stable version  version (2.9.6) see below
      OM+RR: better handling of multiparticles with both photon and massive vector when
         the user asks for longitudinal polarization. Longitudinal photons are automatically
	 discarded. (This overwrites the LTS fix that was simply making the code crash)
      OM: Fixing a bug introduced in 3.1.0, where the automatic setup of FxFx
          was broken.
      OM: Code update to avoid randomness in ordering in code generation (only in debug mode)	  
         

3.2.0 (22/08/21)
      SF/OM/MZ/XZ: Implementation (at LO) of ISR and beamstrahlung for e+e- colliders
      BUG FIX (from 2.9.5 long term stable, see below)
      OM: Fix an infinite loop when running with condor cluster
          or with other cluster/multi-core when cluster_temp_path
	  was used.
      
3.1.1 (28/05/21)
      ALL: put a crash for potentially ambiguous syntax with explanation
           on how to bypass such crash via a new entry in
	   input/mg5_configuration.txt
      BUG FIX (from 2.9.4 long term stable)

3.1.0 (30/03/21)
    ALL: include all changes up to 2.9.3 of the 2.x release
    ALL: pass to python3 by default
    MZ+RF+OM: Fix relevant to the case when PDG specific cuts are set in the NLO run_card.
        The generation of tau was not taking into account the information.
    DP+HS+MZ: Allow for the computation of NLO EW and complete NLO corrections in the 4FS
    MZ+SC+ERN+CS: The code can be linked to PineAPPL (arXiv:2008.12789), making it possible
	 to generate PDF-independent fast-interpolation grids including EW corrections.
	 This supersedes the interface to ApplGrid+aMCFast which was not working in v3
    MZ: Change syntax meaning in 3.0 series to be consistent with the one of 2.0 series:
        QCD<=X, QED<=Y apply at the amplitude level (not to the squared amplitude anymore).
        Squared-amplitude constraints can be imposed by using the usual squared-order 
        syntax, e.g.:
	QCD^2<=X, QED^2<=Y (constrained at the diagram squared level).
    MZ: Different coupling combinations and their scale variations are written in the
        event file as extra weights
    OM: Some aliases have been added for the orders constraints:
            - aEW<=X / aS<=Y is equivalent to QCD^2<=2X / QED^2<=2Y
            - EW<=X / EW^2<=X is equivalent to QED<=X / QED^2<=X

                        ** Long Term Stable Update **


2.9.27 (05/01/26): LAST Bug fix for that LTS version
   OM+RF: Fix an issue where overweights were discarded, leading to some under-estimation
          of the tail (and of the error) 
   OM: fix madspin onshell in the computation of the BR when multiple decays were present
   OM: correct an issue with exrootanalysis where the event file was deleted due to an unhandled flag on how the file was unzipped
   OM: add warning message for python37,3.8,3.9
   Alexander Puck Neuwirth: Fix quotes and permissions in PYTHIA8 script

2.9.26 (30/10/25):
  OM: setup the code to not print the python3.12 syntax warning anymore (still printed in debug mode)
  OM: fix card edition issue when combining "set" and "edit" command
  OM: fix some additional issues related to GCC15


2.9.25 (11/9/25):
  OM: (thanks to SungBeom Cho) crash when running multi_run during the merging of multiple lhe files if
      MadSpin was activated (if the BR was different from one). If the BR=1, the merging was going through but
      the merged sample would contain both decayed and undecayed events, making that file pointless.
      (fixed first in 3.5.9)
  OM (thanks to Julien for the help) fixing a wrong color computation in the presence of sextets where the color matrix
     was wrong starting at 2>4

2.9.24 (23/05/25)
    OM: Fix some compilation issues with the GCC15 compiler
    OM: Fix a serious bug in MadSpin where the benchmark used to generate the sample was ignored by MadSpin and it was using the default of the benchmark instead. This was clearly indicated by a warning stating the fact that the default benchmark was used.
        This bug is more compiler specific than madgraph specific, so this bug is likely present since MG5aMC v2.0.0 if you use a recent version of GCC (like GCC13 or GCC14)

2.9.23 (14/03/25)
    MZ: fix in the pseudorapidity function in lhe_parser FourMomentum class (super minor issue) 

2.9.22 (29/11/24):
    OM: Fix an (NLO) issue if the environment variable FFLAGS was set
    OM: set auto_update X, now automatically set the configuration file (as it should).
    OM: Fix a bug in helicity recycling with scan for rare case where a matrix element is first correct, then zero, then has a single helicity.
        In that third case, the code was not recompiled correctly and was picking the wrong amplitude.
        Thanks to Matteo Maltoni for finding such bug
    OM: Starting support for python3.13, as for python3.12, functionality like reweighting will not work with the 2.9.x LTS branch.

2.9.21 (26/9/24):
    OM: Adding support for GCC14 (thanks to joequant)
    OM: Fixing a rare issue within aloha where some generated code was not compiling
    OM: Fix issue with Pythia8 when running in run_mode=0
    OM: Fixing gnuplot related issue (when pythia8 was running)
    OM: Avoid a numerical inaccuracy issue when the user requested a boost to the center of mass frame for a non Lorentz invariant amplitude
    OM: Avoid to use sde=2 by default for more processes (like Drell-Yan)
    
2.9.20 (17/6/24):
    OM: Fix a wrong syntax handling for decay chain like
        generate ... > A A B  , B > ,  A >  , A >
        where due to the different order in production (AAB) and in decay (BAA)
        all A particles were decayed with both decay modes, instead of with their specific ones.
        (present since v1.0.0)
        Note this bug can still be present if A or B are multi-particles, in that case we advise to use
        the same order for the production and decay syntax.
    OM: Fix an issue with helicity recycling if some processes did not have available phase-space.
        In some situations, the cross-section was reported as simply zero even if other processes were
        kinematically allowed (thanks Sihyun Jeon)
    OM: Improving the detection of cache invalidity for UFO model (thanks A. Valassi)
    OM: Fix compilation of Stdhep on some unix machines (thanks jmcarcell)
    OM: Fix an issue with the weight normalization when running PY8@LO


2.9.19 (20/03/24):
    OM: Fix a "cot" issue introduced in 2.9.18.
    OM: Fixing some regular expressions for the way the fortran writer works.
        The bugged expression was not spotted as problematic for the LTS version.
    OM: Fix some compiler issues with modern compilers for rambo (standalone mode)
    OM: Fix one rare bug for helicity recycling (case without propagator)

2.9.18 (08/12/23):
    OM: Fix a bug for processes extremely close to threshold, where
        the initial grid for some angle was wrongly biased towards \theta=\pi.
        In extreme case, the determination of the non-zero helicity was starting to 
        remove helicities proportional to (1+\cos\theta) since theta was so close to \pi.
        While the grid bias was corrected after the first iteration, the accidental removal of helicity
        could lead to a bias in the cross-section. One way to detect if you are impacted by such a bug is to
        run 5 times with different seeds: if the cross-section fluctuates between two values then you are
        impacted by such issue.
    OM: Implement better handling of the phase-space when you have a S-channel Breit-Wigner but where cuts prevent
        such a resonance from being onshell. (Thanks Sihyun Jeon)
    OM: Implement better handling of the MMNL cut if the process has exactly two leptons (including neutrino).
        This improves the phase-space integrator when MMNL is used for putting invariant mass cut on single W production
    OM: Change the support of the function "cot" within ufo model to have it consistent with UFO convention. 
        It might potentially impact old UFO models; please contact us if this is the case    
    OM: Fix a critical bug when exporting code with two different models where the second model was not correctly imported
        (likely a python3 only bug)
    OM: Fixing issue when using cluster_tmp_dir in multicore mode    
    OM: add a warning concerning the status with python3.12        

2.9.17(19/10/23):
    OM: change integration strategy (SDE=1) for width computation
    OM: clearer error message in some situation

2.9.16(26/07/23):
     OM: fix for helicity recycling
     OM: Update numpy support to latest version
     OM: Fix some interface issue

2.9.15 (12/05/23)
     OM: Fix a bug where the number of cores used was higher than requested
     OM: Reduce RAM used by madspin
     OM: Fix a rare issue with helicity recycling leading to an incorrect format in the generated code (and a crash)
     OM: Fix an issue with param_card parsing for some blocks not related to the model
     OM: Fix compilation error for stdhep
     OM: Fix one issue for sextet model/process
     
2.9.14 (17/02/2023)
     OM: Fix some "output aloha" commands, where some routines were not defined
     OM: Fix a crash in the multicore mode when 0 job was needed. (introduced in 2.9.13)

2.9.13 (01/12/2022)
     OM: Fixing display diagrams on MacOS Catalina
     OM: fixing issue within reweighting module where runname was not respected during scan
     OM: remove unused thread which should allow large scan
     OM: fix a compilation issue of IREGI
     OM: fix auto-download of pdf
     OM: fix systematics.py which was not called with the most appropriate default for matching/merging generation (MLM)

2.9.12 (09/08/2022)
     OM: Fixing an issue with models (like SMEFTatNLO) where some interactions were using more than 9 couplings.
         The couplings were assigned to the wrong Lorentz structure
     OM: Fixing another issue for BSM models where an interaction is associated
         to many couplings (detected for 10 couplings) where helicity recycling
	 was discarding incorrectly such type of interaction.
     OM: Fixing an helicity recycling issue if during a benchmark scan a full
         matrix-element did not have any contribution (all amplitude set to zero)
	 but not in the following one. Those two runs were fine but any
	 following points in the scan will automatically discard that
	 matrix-element.

2.9.11 (03/06/22)
     OM: Fixing a bug on the color assignment for b-quark (that's leading to a crash of herwig)
         This occurs in five flavor models, if you do not force yourself the definition of the
	 multi-particle "j" and "b". This bug was present since 2.3.1 (where the flipping of such
	 multi-particle to 4/5 flavor was done automatically for the first time)
	 
2.9.10 (06/05/2022)
     OM: allow model with GC in their name to use helicity recycling
     OM: Forbid madspin to run with crazy values of BWcut
     OM: Fixing some IO issue with Madspin that was leading to a lock of the code
     OM: set a user interface for the set ewscheme option 
     OM: forbid helicity recycling for models with spin2 and spin3/2 (optimization is not implemented for such case)

2.9.9 (25/02/2022)
     OM: Fix a bug introduced in 2.9.0 for MLM generation in the presence of mixed EW/QCD processes.
         The bug typically leads to a crash within the systematics.py due to wrong power of alpha_s
	 The bug might sometimes not lead to a crash but the impact is "limited" to a change in the (various)
	 scale choice associated to the events. Therefore this is likely to be within scale uncertainty. It can
	 impact the matching with the parton-shower (but this should be visible within the DJR validation plot). 
     OM: Fix some wrong zero results that occur in the presence of conflicting Breit-Wigners.
     OM: Fix an issue when using "$ X" when X~ can be onshell, the phase-space symmetry factor was wrongly set.

2.9.8 (21/02/2022)
     OM: Fix in madspin where onlyhelicity mode was not working anymore
         also allows to not specify any decay in that mode.
     OM: Fix in multi_run mode where some meta-data were incorrectly set within the merged lhef file.
         This occurs ONLY in the presence of non positive definite cross-section. 
         Only the normalization of the cross-section/weight/statistical error could be wrong. Shapes are not impacted.

2.9.7 (29/11/21)
     OM: Fix the behavior of python seed for madevent/madwidth/mcatnlo/madspin
         Now the first of those to set up a python seed will forbid any future reset of the seed
	 Frequent reset of the seed when moving to one package to the next was creating
	 a bias in the effective branching ratio out of madspin (observed only of NLO sample)
	 see: https://bugs.launchpad.net/mg5amcnlo/+bug/1951120
	 Thanks Hannes for the information.
     RF+PT+SF: Fixed a 10-year-old bug that was there in the importance sampling over the FKS configurations.
	 - Fix to pick one shower scale for the S event at random among the FKS configurations,
	   instead of taking the weighted average.
	 - removed the n-body contributions from the random picking of the showerscale among FKS configurations.
        Those bugs were leading to an incorrect pick of the shower scale. The observed impact is relatively small
	and occur in the matching region.

2.9.6 (02/11/21)
     OM: Forbid the possibility to ask for massless bosons to be longitudinally polarised.
         Asking for those with a large multi-particle label could have hidden the fact that the code
	 was returning a non zero cross-section for such request.
     OM: for process like p p > w+{X} w+{Y}, w+ > l+ vl
         i.e. processes with polarization of identical particles where the decay involves multi-particles and
         where the leptons do not all have the same mass, some of the processes were incorrectly discarded.
     OM: fix an issue for photon initial state with lpp=+-4 where the lhef output was not setting the muon
         as the correct colliding particles (thanks to Yuunjia Bao) 	  
     OM: Fixing an issue that madspin was not always using the python executable that was used by the
         mg5_aMC executable
     OM+PT: Change the shower script running pythia8 in order to be working with pythia8.3.
         Running with Pythia8.2 (or lower) is then not possible anymore.
     OM: Fix issue that some loop-induced processes were not working anymore since 2.9.0
     OM: Fix the default set of cuts present in the default run_card if gluons were present in the final state but no
         light (or b) quark. The absence of those cuts, in that situation, was leading to hardcoded cuts that were
	 not easy to overwrite for a non expert user.
	 
2.9.5 (22/08/21)
      OM+LM: [LO only] Fix the factorization scale dependence for lpp=2/3/4.
            This was claimed to be using a fixed scale computation while
	    in some cases the scale was dynamical
            To have full flexibility we introduced two additional
	    (hidden) parameters: "fixed_fac_scale1" and "fixed_fac_scale2"
	    that allow to choose fixed scale for only one beam.
      OM: fix auto-width that was not following	run_mode specification
      OM: fixing missing rwgt_info for reweighting with gridpack mode
      OM: fixing 'check' command with the skip_event mode
      OM: fixing auto-width computation for 3 body decays and identical particles which was sometimes leading to a crash
      OM: Fix some potential infinite loop when running with python3
      
2.9.4(30/05/21)
      OM: Fix a python3 issue for madSpin when using in a gridpack mode (set ms_dir)
      OM: Fix an issue for non positive definite matrix-element when using the
          "set group_subprocesses False"  mode of MG5aMc
      OM: Fix a speed issue for gridpacks in readonly mode where the grids were recomputed



                        ** PARALLEL VERSION FOR EW branch **

3.0.3 (06/07/20)
     include up to 2.7.3

3.0.2
     include up to 2.7.2
     MZ: better handling of zeros in color-linked borns 
     RF: improved the behaviour for integration channels that give zero cross section
     RF: Rewritten the mint integrator package. Makes for easier extensions.
     RF: Fixed a problem related to the multi-channel enhancement factor for
         NLO runs that only appeared in BSM theories (or EFTs)
     RF: In the virt-tricks, replaced the grids for the virtuals with a
         polynomial fit, which is more precise and allows for a reduction of
         the number of calls to the virtual. However, requires to save all the
         phase-space points for which the virtual has been computed to disk,
         hence requires more disk space and memory for highly-accurate
         runs. Can be disabled in the FKS_params.dat card, if disk space
         and/or memory use needs to be limited.

3.0.1 
     include up to 2.6.4
     MZ: Enable shower for QCD-only splittings
     RF: Fixed a major bug in the code that affected 3.0.0, which affected processes with
         cuts and --to a lesser extent-- processes with massive final state particles:
	 the phase-space region where "xi_i_fks != xinorm*xi_i_hat" had the wrong "prefact".

3.0.0 (01/05/18):
     RF+SF+VH+DP+HS+MZ: Allow for the computation of NLO EW corrections (and more subleading NLO corrections).
         [Note: when requiring NLO corrections ([QCD], [QED] or [QCD QED]), internally the code always treats this [QCD QED],
	 and then uses the coupling orders to specify which contributions should be considered. Among other things, this means
	 that when doing QCD corrections only --contrary to previous versions-- also loop diagrams with non-QCD-charged 
	 particles will be included (e.g., the pentagons in Higgs production via vector-boson fusion).]
     HS: inclusion of SMWidth for fast evaluation of NLO accurate widths for all SM particles
     RF+SF: Resonance-aware phase-space mapping for NLO computations 
     RF: Reduction of the number of jobs -- results in earlier detection of small integration channels for fNLO
     OM: remove ./bin/mg5 executable

                             ** Main Branch (version 2.x) Update **


2.9.3(25/03/21)
      OM: Fix an issue with t-channel particles for massive initial states where sqrt(S)
          was close to the initial mass. Boundary conditions on t-channel were
	  incorrectly set up, leading to strange spectra in various distributions (very old bug)
      OM: Fix an issue with a crash for loop computation
      OM: more python3 fixing bug
      OM: Fix a bug in the re-writing of the run_card when the seed was specified, leading to a wrong templating
      OM: various small fixes of crashes/better logging/debug information
     
2.9.2(14/02/21)
      MZ+RF+OM: Fix relevant to the case when PDG specific cuts are set in the NLO run_card.
        The generation of tau was not taking into account the information.
      OM: fix an issue when running with python3 where the normalization of pythia8 files was wrong
          when pythia8 was run in parallel.
      OM: fixing more issues related to MSSM (introduced within 2.9.0)
      OM: fixing an issue with loop-induced processes (introduced within 2.9.0)
      OM: Fix some wrong scale variation at NLO+PS (introduced in 2.9.0)
      OM: Fix an issue with the installation of pythia-pgs
      

2.9.1.2 (02/02/21)
      Kiran+OM: Fixing issue for the MSSM model
      OM: Fixing issue with mac support
      OM: fix a bug introduced within 2.8.2 related to a wrong phase-space mapping of conflicting resonances.
          Leading to zero cross-section and/or a potentially biased cross-section for more complex cases.
	  The issue occurs only if you have a conflicting resonance followed by a massless propagator.
	  one example is generate p p > z h , z > e+ e-, h > b b~ a

2.9.0 (30/01/21)    **** Major speed-up update for LO computation ***
      Kiran+OM: Optimization of the matrix-element at run-time using recycling helicity method
                - This speeds up LO computation by a factor of around 2
		- This can be turned off via the command "output PATH --hel_recycling=False"
      OM: Various optimizations for T-channel integration
          - use different orderings for T-channel. Four different orderings have been implemented
	    at generation time, the code decides which ordering to use channel-per-channel.
	    This can be turned off via the command "output PATH --t_strategy=2"
	  - increase number of maximum iteration for double or more T-channel
	  - implement an alternative to the standard multi-channel.
	     instead of using the amplitude squared it uses the product of the denominators
	     (the default strategy use depends of the process)
	  - speed-ups for VBF type processes are often around 100 times faster
	  - This fixes a lot of issues for processes failing to generate the targeted number of events.
      OM: Better version of the color computation (this leads to only a modest gain except for very complex processes).
          This optimization leads to a longer generation time of the code (only for complex processes).
	  This can be turned off  via "output PATH --jamp_optim=False"
      OM: New parameter in the run_card:
          - sde_strategy: allows to change the default multi-channel strategy
	  - new phase-space optimization parameters are now easily available via the command "update ps_optim" 
      OM: New global parameter "auto_convert_model"
          if set on True (set auto_convert_model T or via input/mg5_configuration.txt)
	  all model crashing due to a python3 compatibility issue of the UFO model will be automatically converted
	  to a python3 compatible model. 

2.8.3 (26/01/21):
      OM: Bunch of bug fixing related to python3 issues (mainly related to unicode encoding)
      OM: Various fix for reweighting with loop (mainly with python3 as well)
      OM: Fix a potential bug for polarized samples with at least three polarised particles, with at least two
          identical particles and one different polarised particle.
      OM: Fix various bugs for the maddm interface (thanks Daniele)
      OM: Fix various issues with overall order usage
      OM: Fix compatibility with MacOS 11
      OM: fix additional GCC10 compatibility issue
      OM: fix an issue for 1>n processes where widths were set to zero automatically (introduced in 2.8.0)
      OM: fix aloha output mode for python output
      OM: avoid a bug with helicity filtering that was kept for all benchmark when using
          multiple successive re-weighting
      OM: Edition of the reweighting card via the "set" command is now starting from the original param_card
          used to generate the events file, and not from the model default anymore.
      OM: Interference terms now have their default dynamical scale set to HT/2


2.8.2 (30/10/20):
      OM: Fix a bug when setting widths to zero where they were actually set to 1e-6 times the width
          This can lead to bias in the cross-section if your process is very sensitive to the cross-section
	  due to gauge cancellation. Bug introduced in version 2.6.4.
      OM: Hide Block parameter loop from NLO model but for madloop run.
          The factorization scale is still determined by the setup of the run_card as before
      OM: Fix a couple of python3 specific bugs

2.8.1(24/09/20):
      OM: Change user interface related to FxFx mode
          - If you have multiple multiplicities at NLO,
	    - the default run_card is modified accordingly
	        - has ickkw=3 and follows the official FxFx recommendation (i.e. for
		  scale and jet algo)
            - the shower card has two parameters that are dynamically set
                - njmax (default: -1) is automatically set depending on the process definition
                - Qcut (default: -1) is now automatically set to twice ptj parameter
            - the default parton-shower is automatically set to Pythia8
          - The value of the cross-section after FxFx formalism (removing double counting) is
	    now printed on the log and available on the HTML page
      OM: Fix for the auto width for three body decays in the presence of identical particles.
      OM: add support for __header__ in UFO model
      OM: allow restriction card to have auto-width
      OM: fixing some html links (removed ajax links forbidden by major web browsers)
      OM: Various fix related to the python3 support
          - including more efficient model conversion method

2.8.0 (21/08/20):
      OM: pass to python3 by default
      OM: For LO processes, you can now set lpp1 and lpp2 to "4" for processes with an initial photon in order to get the
          effective photon approximation. This mode behaves like the "3" one: The cut-off scale of the approximation
	  is taken from the fixed renormalization factorisation scale: "dsqrt_q2fact1/dsqrt_q2fact2"
      OM: The widths of T-channel propagators are now set to zero automatically.
          To return to the previous behaviour, you can use the option "set zerowidth_tchannel False"
	  (to set before the output command)
      OM: Change in madevent phase-space integrator for T-channel:
            - The integration of photon/Z/Higgs is now done together rather than separately and follows importance
	       sampling of the photon channel
	    - A new option "set max_t_for_channel X" allows to veto some channel of integration with more than X
	       t-channel propagators. This option can significantly speed up the computation in VBF processes.
	       (We advise to set X to 2 in those cases)
	    - Fix a numerical issue occurring for low invariant mass in T-channel, creating spurious configurations.
      OM: In madspin_card you can now replace the line "launch"
          by "launch -n NAME", this will allow to specify the name of the
	  directory in EVENTS where that run is stored.
      OM: Change in the python interface of the standalone output. On top of the pdg of the particles,
          you can now use the process_id (the one specified with @X) to distinguish processes with the same particle
	  content. This parameter can be set to -1 and the function then ignores that parameter.
      OM: Adding new option in reweighting to allow the user to use the process_id in the presence of ambiguous
          initial/final state.
      OM: Update the makefile of standalone interface to python to be able to compile in multicore.
           (thanks Matthias Komm)
      OM: Update of the auto-width code to support UFO form-factors
      OM: Fixing a numerical issue with the boost in EPA mode.
     

                        ** PARALLEL VERSION FOR PYTHON 3 **

2.7.3.py3(28/06/20):
      ALL: Contains all features of 2.7.3 (see below)
      OM: Fix a crash when running PY8 in matched/merged mode (bug not present in not .py3 version of the code)
      MZ: low_mem_multicore_nlo_generation is working again
          but not for LOonly mode and not in OLP mode

2.7.2.py3(25/03/20):
      ALL: Contains all features of 2.7.2 (see below)

2.7.1.py3(09/03/20):
      ALL: Contains all features of 2.7.1 (see below)
      OM: Fixed a lot of python3 compatibility issues raised by users
          Particular Thanks to Congqiao Li (CMS), Richard Ruiz and Leif Gellersen
	  for their reports.

2.7.0.py3 (27/02/20):
      ALL: Contains all features of 2.7.0
      OM: Support for python3.7 on top of python 2.7
          - python2.6  is not supported anymore
	  - this requires the module "six" [pip install six --user]
      OM: dropping function set low_mem_multicore_nlo
      OM: dropping support for syscalc (c++ version)
      OM: introduction of new setup variable
          - f2py_compiler_py2 and f2py_compiler_py3
	    which will be used to overwrite f2py_compiler when using the associate python version
          - lhapdf_py2 and lhapdf_py3
	    same for lhapdf
      OM: introduction of a new command "convert model FULLPATH"
          - try to convert a UFO model compatible with python2 only to a new model compatible both with
	    Python2 and Python3 (no guarantee)


                             ** Main Branch  Update **

2.7.3(21/06/20)
      OM: Fixing some bugs for read-only LO gridpacks (wrong cross-section and shape when generating events).
          Thanks to Congqiao Li for this
      OM: Allowing loop-induced process to run on LO gridpack with read-only mode.
          Thanks to Congqiao Li for this      
      OM: Fix a bug in the longitudinal polarization for off-shell effect, leading to deviation at large invariant mass.
      OM: Adding more option to the run_card for fine tuning phase-space integration steps
          All are hidden by default:
	    - hard_survey [default=1]: request for more points in survey (and subsequent refine)
	    - second_refine_treshold [default=1.5]: forbid the second refine if the cross section after the first refine
	       is smaller than the cross-section of the survey times such a threshold
      OM: new commands for the edition of the cards:
           - set nodecay: remove all decay line from the madspin_card
	   - set BLOCKNAME all VALUE: set all entries of the param_card "BLOCKNAME" to VALUE
	   - edit CARDNAME --comment_line='<regular_expression>' : new syntax to comment all lines of a card
	       that are matching a given regular expression
      OM: For Mac only, when running in script mode, MG5aMC will now prevent the computer from going to idle sleep.
          You can prevent this by running with the '-s' option. like ./bin/mg5_aMC -s PATH_TO_CMD


2.7.2(17/03/20)
      OM: Fix a bug in pythia8 running on Ubuntu 18.04.4 machines
      OM: Speed up standalone_cpp code by changing compilation flag

2.7.1.2(09/03/20)
      OM: Fixing issue (wrong cross-section and differential cross-section) for
          polarised sample when
	   1) you have identical polarised particles
	   2) those particles are decayed
	  examples: p p > j j w+{0} w+{T}, w+ > e+ e-
      OM: In the presence of identical particles, if you define the exact number of decays
          (either via the decay chain syntax or via MadSpin) then they are assigned in an ordered way:
          generate p p > z{0} z{T}, z > l+ l-, z > j j
	  means that the longitudinal Z decays to leptons (and the transverse one to jets)
	  Thanks to Jie Xiao for reporting such issues. 
      OM: The Effective Photon Approximation is now always done with a fixed cutoff. This uses the FIXED factorization scale
          of the associate beam as the cutoff of the Improved Weizsaecker-Williams.
      OM: Allow to install lhapdf6.2 by default
      OM: Fixing the website used to download pdfs, since lhapdf removed their previous hepforge page for pdf sets.
      OM: Fixed a bug (leading to a crash) introduced in 2.6.6 related to the ckkw/MLM check for BSM models
          with gluons with non QCD interactions
      OM: For fortran standalone with python binding, it is not necessary anymore to run the code from a specific directory.
          You do however need to use the standard path for the param_card or have the file ident_card within the same
	  directory as the param_card.dat.


2.7.0(20/01/20)
      OM: Allow for a new syntax in the presence of multi-jet/lepton processes:
         generate p p > 3j replaces p p > j j j 
      OM: Allow syntax for (fully) polarized particles at LO: [1912.01725]
           ex: p p > w+{0} j, e+{L} e-{R} > mu+ mu- Z{T}
               p p > w+{T} w-{0} j j, w+ > e+ ve, w- > j j
      OM: (Thanks to K. Mawatari, K. Hagiwara) implementation of the axial gauge for the photon/gluon propagator.
          via "set gauge axial"
      OM: Support for elastic photons from heavy ions implemented. For PA collisions, you have to generate your diagrams
          with "set group_subprocesses False"
      OM: The default run_card.dat is now by default even more specific to your process. 
          Nearly all the cuts are now hidden by default if they do not impact your current process.
          This allows to have less information by default in the run_card, which should simplify its
          readability.
      OM: distinguish in the code if zero contributions are related to no point passing cuts or if
          they are related to a vanishing matrix-element. In the latter case, allow for a lower threshold.
          (This allows to speed up the computation of such zero contributions)


2.6.7(16/10/19)
      OM: Fix a bug introduced in 2.6.2, some processes with gluon-like particles which can lead to the wrong sign for the interference term.
      OM: Fix a bug introduced in 2.6.6 related to the restriction of models which was leading to a wrong result for re-weighting with loop models (but impact can in principle be not limited to re-weighting).
      OM: systematics now supports the option --weight_format and --weight_info (see help command for details)
      OM: set the auto_ptj_mjj variable to True by default
      OM: the systematics_arguments default value is modified in the presence of matching/merging.
      OM: reweight: add an option --rwgt_info to allow to customise the banner information associated to that weight
      RF: Fixed a bug in the warning when using FxFx in conjunction with Herwig++/Herwig7. Also, with latest version of
          Herwig7.1.x, the FxFx needed files are compiled by default within the Herwig code. Thanks Andreas Papaefstathiou.

2.6.6(28/07/19)
      OM: Bug in the edition of the shower_card. The set command for logical parameters was never setting the
          parameter to False. (Thanks to Richard Ruiz) 
      RF: Fixed a bug in the creation of the energy-stripped i_FKS momentum. Tested for several processes,
          and seems to have been completely harmless.
      OM: Forbidding the use of CKKW/default scale for some UFO models allowing gluon emission from quarks with
          no dependence on aS
      OM: Add an option "keep_ordering" for reweighting feature to allow to sometimes use decay chain even if
          you have ambiguity between final state particles.
      OM: Fixed an issue with interference which sometimes happens when some polarization contributions were negative but not all of them.
      VH: Change in the pythia8 output mode (thanks to Peter Skands)
      OM: Any number in the cards (not only integers) can use multiplication, division and k/M suffix for times 1000 and 1 million respectively
      OM: Energy cuts (at LO) are now hidden by default for LHC type runs but visible for lepton collider ones.
      OM: Same for beam polarization

2.6.5 (03/02/19)
      OM: Fix some speed issues with the generated gridpack --speed issue introduced in 2.6.1--
      OM: Fix a bug in the computation of systematics when running with Python 2.6.
      OM: import model PATH, where PATH does not exist yet, will now connect to the online db
          if the model_name is present in the online db, then the model will be installed in the specified path.
      MZ: Applgrid+aMCFast was broken for some processes (since 2.6.0), due to wrong 
          information written into initial_states_map.dat. This has been fixed now
      OM: change in the gridpack. It automatically runs systematics.py (if configured in the run_card)
      OM: Fix a MLM crash occurring for p p > go go (0,1,2 j)
      OM: Fix an issue for BSM models with an additional colored particle where the default dynamical scale choice
          was crashing


2.6.4 (09/11/18)
      OM: add specific treatment for small width (at LO only and not for loop-induced)
          if the width is smaller than 1e-6 times the mass, a fake width (at that value) is used for the
          numerical evaluation of the matrix-element. S-channel resonances are re-scaled according to 
          narrow-width approximation to return the correct total cross-section (the distribution of events 
          will on the other hand follow the new width). 
          The parameter '1e-6' can be changed by adding to the (LO) run_card the parameter: "small_width_treatment"
      OM: add a new command "install looptools" to trigger the question that is automatically triggered
          the first time a loop computation is needed.
      RF: Fixed a bug when using TopDrawer plots for f(N)LO runs, where the combination of the plots could lead
          to completely wrong histograms/distributions in case of high-precision runs.
      OM: Fix some MLM crashes for some processes (in particular BSM processes with W').
      OM: Fix a bug in the reweighting due to the new lhe format (the one avoiding some issue with py8)
      OM: Fix a behavior for negative mass, the width was set to negative in the param_card automatically
          making the parton shower (and other code) crash since this does not follow the convention.
      OM: Change compiler flag to support Mojave.

2.6.3.2 (22/06/18)
      OM: Fix a bug in auto-width when masses are below the QCD scale.
      OM: Fix a bug for g b initial state where the mass in the lhe file was not always correctly assigned
          Note that the momentum was fine (i.e. in the file P^2 was not equal to the mentioned M but to the correct one)
      OM: Improvement for madspin in the mode spinmode=none
      OM: Fix a bug in MadSpin which was making MadSpin work only in debug mode

2.6.3 (15/06/18)
      OM: When importing model, we now run one additional layer of optimisation:
           - if a vertex has two identical couplings for the same color structure then the associated Lorentz
             structures are merged into a single one and the vertex is modified accordingly
      OM: When restricting a model, we also run one additional layer of optimisation
           - Opposite sign couplings are now identified and merged into a single one
           - if a vertex has two identical couplings (up to the sign) for the same color structure
             then the associated Lorentz structures are merged into a single one and the
             vertex is modified accordingly 
      VH+OM: changing the ALOHA naming scheme for combined routines when the function name becomes too long.
      OM: adding a hidden parameter to the run_card (python_seed) to allow to control the random number 
          generated within python and be able to have full reproducibility of the events
      OM: Fixing some issue with the default dynamical scale choice for 
            - non minimal QED sample
            - the heft model when there is multiple radiation coming from the Higgs decay/scattering
          This can also impact MLM since it uses the same definition for the dynamical scale
      OM: Fix some issue for DIS scattering where the shat was wrongly defined for low energy scattering.
          Low energy scatterings are not advised since they break the factorization theorem.
          In particular the z-boost of the events is quite ill defined in that scenario.
      OM: changing the format of the param_card for NLO model to match expectation from the latest PY8
      OM: Update of MadSpin to allow special input file for the case of spinmode=none. 
          With that very simple mode of decay, you can now decay hepmc files or wrongly formatted leshouches events
          (in that mode we do not have spin correlation and width effects)
      PT: in montecarlocounter.f: improved colour-flow treatment in the case gluons are twice colour-connected to each other
          new gfunction(w) to get smoothly to 0 as w -> 1. (for NLO+PS run)
      OM: Fix some issues for the new QED model (including one in the handling of complex mass scheme of such model)
      OM: Fixing an issue of the param_card out of sync when running compute-widths
      OM: Adding a Qnumbers block for ghosts (the latest version of py8 was crashing due to their absence)

2.6.2 (29/04/18)

      Heavy ion pdf / pdf in general:
      -------------------------------
      OM: Support for rescaling PDF to ion PDF (assuming independent hadron), this is well suited for Lead-Lead collisions, p-Lead collisions and fixed-target
      OM: Support in systematics.py for ion pdf. Possibility to rescale only one beam (useful to change only one PDF for fixed-target experiments)
      OM: Removing internal support for old type of PDF (only supported internal pdf are now cteq6 and nnpdf23)


      User Interface
      --------------
      OM: introduce the "update to_full" command to display all the hidden parameters.
      OM: introduce "update ion_pdf" and "update beam_pol" to add related section in the run_card.
          the polarization of the beam is set as a hidden parameter instead of as a default parameter
      OM: improve handling of (some) run_card parameters:
            - add comment that can be displayed via "help NAME"
            - add autocompletion for some parameters
            - add direct rejection of parameter edition if not in some allowed list/range
      
      Bug fixes:
      ----------
      OM: Fix issue with madspin introduced in 2.6.1 (failing to generate correct diagram for NLO process)
      OM: fix crash in 1>N reweighting
      RF+MZ: Fixed a problem with (f)NLO(+PS) runs in case the Born has identical QCD-charged
      	     particles. Cross sections were typically correct, but some distributions might
	     have shown an asymmetry. 
      OM: Change in LO matching for HEFT (or any model with an hgg vertex) in the way to flag jets that should
          not take part in the matching/merging procedure.
      OM: Fixed a bug for loop induced in gridpack mode
      RF: Fixed a bug for ApplGrid: in rare cases the ApplGrid tables were filled twice for the same event
      OM: Fixed a bug for fixed target experiment when the energy of the beam was set to 0. 
      RF: Fixed an issue where too many files were opened for fNLO runs in rare cases
      OM: Fix an issue in the MadEvent html output where some links were broken
      OM: Fix an issue for the display lorentz function (which was also presenting a security issue for online use)
      OM: Fix issues for spin 3/2 (one in the presence of fermion flow violation and one for custom propagators)

      Enhancement:
      -----------
      OM: add the value of all the widths in Auto-width in the scan summary file
      OM: For 1>N, if the user sets fixed_run_scale to True, then the scale is chosen accordingly
          and not following the mass of the initial state anymore
      RF: For the HwU histograms, if no gnuplot installation is found, write the gnuplot scripts in v5
          format (instead of the very old v4 format).
      OM: Change the default LO output directory structure. Now by default the lepton and neutrino are split 
          in two different directories. This avoids facing problems with the asymmetric cuts on lepton/neutrino
      OM: loop-filter commands are now working for loop-induced processes
      OM: New method avoiding that two processes run inside the same output directory.
          This is implemented only for Gridpack and LO run so far. 
          The new method should be more robust in case of a crash (i.e. not wrongly triggered as before)
      OM: For LO scan, if a crash (or ctrl-c) occurs during the scan, the original param_card is now
          restored.

2.6.1 (12/12/17)

      RF+MZ: It is now possible to add LO matrix elements (with [LOonly]) 
             to Fx-Fx merged samples. Thanks to Josh Bendavid for testing.
      OM: Re-factoring the code asking which program to use (both at LO and NLO)
          - design modular, designed for PLUGIN interactions
          - the length of the question auto adapts to the size of the shell
      OM: Allowing to have the gridpack stored on a readonly filesystem
      OM: Fix a bug in matching/merging forbidding the pdf reweighting for some processes (since 2.4.0)
      OM: Creation of an online database with the names of known UFO models. If a user tries to import a model
          which does not exist locally, the code will automatically check that database and download the 
          associate model if it exists. You can contact us if you are the author of one model which is not
          on our database. 
	  The list of all available models is available by typing "display model_list"
      OM: Model with __arxiv__ attribute will display "please cite XXXX.XXXXX when using this model" when
          loaded for the first time.
      OM: A failure to import a UFO model does not try to import a v4 model anymore
      OM: Many model present in models directory have been removed, however they can still be imported
          since they are available via automatic-download
      OM: Refactoring of the gridpack functionality with an infinite loop to reach the requested number of events
      OM+RF: Adding new class of cut at LO/NLO defined via the pdg of the particle
      VH: Support for the latest version of MA5
      MZ: Adding support for lhapdf v6.2
      OM: Fixing various bugs in the spinmode=onshell mode of MadSpin
      OM: Fix a bug for models with 4 fermions in the presence of restrict_card
      OM: Fix an aloha bug in the presence of complex form-factors.
      OM: improve auto-detection and handling of slha1/slha2 input file when expecting slha2. 

2.6.0 (16/08/17)
      New Functionalities:
      --------------------
      RF+OM: Added the possibility to also have a bias-function for event generation at (f)NLO(+PS)
      OM: Improve Re-WEIGHTING module
          1) creation of a single library per hypothesis.
          2) libraries for new hypotheses can be specified via the new
             options: change tree_path and change virt_path
          3) allows to re-weight with different masses in the final states (LO only)
             This forces to rewrite a new lhe file, not adding weight inside the file
             (via the command: change output 2.0)
          4) allows to run the systematics on the newly generated file (for new output file)
             (via the command change systematics True)
          5) Fix some NaN issues for NLO reweighting in the presence of collinear emission
	  6) various bug fixing, speed improvement,...
      OM: Add new option to SYSTEMATICS program:
           --remove_weights, --keep_weights, --start_id
           See "help systematics for more details."
      OM: Allow to specify param_card, run_card,... directly via an html link.

      Bug Fixing:
      -----------
      OM: Update condor class to support CERN type of cluster (thanks Daria Satco)
      OM: Fixing a bug leading to a crash in pythia8 due to one event wrongly written when splitting the
          events for the parallelization of pythia8
      OM: Fixing an issue, leading to NAN for some of the channel of integration for complicated processes.
      OM: Fix a bug in gridpack@NLO which forbids running it when the gridpack was generated with 0 events.
      RF: Fixed bug #1694548 (problem with NLO for QCD-charged heavy vector bosons).
      RF: Another fix (adding on a fix in 2.5.5) related to FxFx merging in case there are
      RF: Fixed bug #1706072 related to wrong path with NLO gridpack mode
          diagrams with 1->3 decays.
      RF: Fixed a bug (found by SF) that gave a seriously biased resonance mass when using MadSpin for decaying a 2->1 process.
      PT: Fix in montecarlocounter.f. Previously, for NLO+PS it was reading some subleading-colour information, now all
          information passed to the MC counterterms is correctly leading colour. 
      OM: Fix systematics computation for lepton collider
      OM+VH: Remove the proposition to install pjfry by default, due to many installation problems. The user can still force
          the installation if they want to.
      OM: Fix a problem of madspin when recomputing width for model loaded with --modelname option
      OM: Fix events writing for DIS (thanks to Sho Iwamoto)
      OM: Fix a problem of output files written .lhe.gz, even if not zipped (python2.6 only)
      OM: Fixing some issues related to the customised propagator options of UFO models
 
      Code Re-factorisation:
      ----------------------
      RF: Refactor of the HwU.f fortran code. Gives more flexibility and potentially lower
          memory requirements.
      RF: Refactor of the (NLO) code related to extra weights (for PDF/scale uncertainties).
      OM: Increase modularity of the code for the support of plugin (maddm and MPI@LO)     
          - Now we support the HPC plugin allowing to generate LO-gridpack on MPI machine
          - Plugin can now use the "launch" keyword

2.5.5(26/05/17)
      OM: Fixing a bug in the creation of the LO gridpack introduced in 2.4.3. Since 2.4.3 the generated
          gridpacks were failing to include the generated grid for each channel. This does not lead to
          bias but to a significant slow down of the associated gridpack.
      OM: Supporting user functions calling other non default functions.
      OM: adding the command "update to_slha1" and "update to_slha2" (still beta)
      RF: some cleanup in the NLO/Template files. Many unused subroutines deleted.
      OM: fixing some bugs related to complex_mass_scheme
      OM: fixing a bug in ALOHA for C++ output (in the presence of form-factors)
      OM: fixing lhe event for 1 to N process such that the <init> block is consistently set for the shower
      OM: ExRootAnalysis interface is modified (need to be requested as an analysis)
      RF: Fix for FxFx merging in case there are diagrams with 1->3 decays.

2.5.4(28/03/17)
      OM: Add a warning in the presence of small widths
      OM: Fix a bug related to a missing library (introduced in 2.5.3)
      OM: Improve stability of the onshell mode of MadSpin
      VH: Fix some problem related to LHAPDF

2.5.3(09/03/17)
      PT: Modified the default shower starting scale in montecarlocounter.f.
          The new reference scale from which the dampening profile is computed is sum_i mt_i/2, i
          being Born level final-state momenta.
      OM: New "special mode" for madspin accessible via "set spinmode onshell".
          This mode allow for full spin-correlation with 3 (or more) body decay but the decaying particle
          remains exactly onshell in this mode. Loop-induced production/decay are not allowed in this mode.
      OM+RF: Allowing for creation of LHE like output of the fixed order run at NLO.
          This LHEF file is invalid for parton-shower (all PS should crash on such file). It will be 
          unphysical to shower such sample anyway.
          Two hidden parameters of the  FO_analyse_card.dat allow some control on the LHEF creation
       	  "fo_lhe_weight_ratio" allows to control the strength of a partial unweighting [default:1e-3]
             increasing this number reduces the LHEF size.
          "fo_lhe_postprocessing" can take value like nogrouping, norandom, noidentification.
             nogrouping forbids the appearance of the LHEF(version2) tag <eventgroup>
             norandom   does not apply the randomization of the events.
             noidentification does not merge born events with other born-like counter-events
      RF: Better job handling for fNLO runs.
      VH: Fixing various problems with the pythia8 interface (especially for MLM merging)
      Team: Fixing a series of small crashes

2.5.2(10/12/16)
      OM: improve systematics (thanks to Philipp Pigard)
      OM: new syntax to modify the run_card: set no_parton_cut
          This removes all the cuts present in the card.
      OM: change the default configuration parameter cluster_local_path to None
      OM: change the syscalc syntax for the pdf to avoid using & since this is not xml compliant
      OM: avoid to bias module to include trivial weight in gridpack mode
      OM: Fix a bug making 2.5.1 not compatible with python2.6
      OM: Improve "add missing" command if a full block is missing
      OM: Fixing a bug reporting wrong cross-section in the lhef <init> flag (only in the presence of 
          more than 80 channels of integration)

2.5.1 (04/11/16)
       PT+MZ: New interface for Herwig7.
              Fixed a bug in the Herwig7/Herwig++ counterterm (relevant to 2 -> n, n>2:
              in particular, the bug affected the dead zone for final-final colour connection
              in processes with more than two particles in the Born final state)
       VH: Parallelization of PY8 at LO
       OM: add the possibility to automatically add missing parameters in a param_card
           with the command "update missing" at the time of the card edition. Useful for
           some SUSY cards where some block entries are sometimes missing.
       OM: Possibility to automatically run the systematics program at NLO, or turn it off at LO
           (hidden entry of the run_card systematics_program = systematics|syscalc|none)
       RF: Some refactoring of the NLO phase-space generation,
           including some small improvements in efficiency.
       OM: Plugins can be included in a directory MG5aMC_PLUGIN; that directory needs to be in
           the $PYTHONPATH
       OM: Fix systematics for e+ e- initial state.
       VH: Fix various bugs in the HepMc handling related to PY8 (LO generation)
       OM: allow to install the maddm functionality (install ./bin/maddm executable)

2.5.0 (08/09/16)
     FUNCTIONALITY
       VH+OM: Adding an official interface to Pythia8 (parton shower) at Leading-Order
	        More information at https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/LOPY8Merging	   
       VH+OM+MA5: Adding an official interface to MadAnalysis5 for plotting/analysis/recasting
	        More information at https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/UseMA5withinMG5aMC	   
       OM: Introduces a new function for LO/NLO interface "systematics"
            This function allows to compute systematics uncertainties from the event sample
            It requires the event sample to have been generated with 
                - use_syst = T (for LO sample)
                - store_reweight_info = T (for NLO sample)
            At LO the code is run automatically if use_syst=T (but only if SysCalc is installed)
       VH+OM: Adding the possibility to bias the event weight for LO generation  via plugin.
            More information: https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/LOEventGenerationBias
       VH+SP: extend support for CKKWL

     CODE IMPROVEMENT / small feature
     -------------------------------
       OM: Modify the structure of the output format such that all the internal format have the same structure
       OM: Adding the Plugin directory. Three kinds of plugin are currently supported
           - plugins defining a new type of output format
           - plugins defining a new type of cluster handling
           - plugins modifying the main interface of MG5aMCnlo
           More informations/examples are available here:
           https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Plugin
       OM: Adding the possibility of having detailed help at the time of the edition of the cards.
            help mass / help mt / help nevents provide some information on the parameters.
       OM: NLO/LO Re-weighting works in multi-core
       OM: add an automatic update of the param_card to write the correct value for all dependent parameters.
       OM: add the check that the param_card is compatible with the model restriction.
       OM: Adding the run_card options "event_norm" for the LO run_card (same meaning as NLO one)
       VH: extend install command to install: lhapdf/pythia8
       VH: Interfaced MadLoop to COLLIER for tensor integral reduction.	   
       OM+VH: At the first loop/NLO computation, a new question will now be asked to choose which program
           to install to compute the loop. You can still install additional method later via the "install" command
       OM: Replace the mssm model by the MSSM_SLHA2 following the SLHA2 instead of the SLHA1 convention

     BUG FIXING
     ----------
       OM: Fix a bug in the helicity by helicity reweighting method. (introduced in 2.4.3)
       OM: Fix a bug in the reweight_card where relative path was not understood from the local directory 
            where the program was run by the user.

2.4.3 (01/08/16)
        OM: Reduce the amount of log file/output generated for LO runs (the output can be up to three times smaller).
        OM: For the LO combination of events (unweighting) pass to the method previously used for loop-induced.
            This method is faster and requires less I/O operations.
            This fully removes the need for the file events.lhe.gz, which is not created anymore (further reducing the output size)
        OM: Optimise the code in order to be able to run scan with more than 2k steps.
        OM: Optimise the lhe_parser module (used for the unweighting/re-weighting/...) around 20% faster than before.
        OM: Fix a bug in MadSpin where the cross-section reported in the <init> block of the LHEF
            was wrongly assigned when multiple processes were present in the LHEF and different branching ratios
            were associated to each of those processes.
        RF: For NLO process generation, fix a problem with turning on PDF reweighting with sets that have only a
            single member. Also, allow for reweighting with up to 25 PDF sets (and their error members) for a single run.
        OM: Fixing a bug allowing to specify a UFO model by its full path for NLO computation (thanks Zachary Marschall).
        OM: Fixing a bug in LO re-weighting in case of helicity by helicity re-weighting. Now the events are boosted back into
            the center of mass frame to ensure consistency with the helicity definition.

2.4.2 (10/06/16)
        OM: fix a compilation problem for non standard gfortran system
        OM: reduce the need for lhapdf for standard LO runs. (was making some runs fail due to missing dependencies)

2.4.1 (10/06/16)
        OM: Fix a bug in fixed-target experiments with PDF on the particle at rest.
            The cross-section was correct but the z-boost was not performed correctly.
        OM: Fix various bugs in MadSpin
        OM: Fix some bugs in MLM merging, where chcluster was forced to True (introduced in 2.2.0)
        OM: Allow to specify a path for a custom directory where to look for model via the environment 
            variable PYTHONPATH. Note that this is used AFTER the standard ./models directory

2.4.0 (12/05/16)
        OM: Allowing the proper NLO reweighting for NLO sample
        RF: For NLO processes allow for multiple PDF and scale reweighting, directly by inputting lists
            in the run_card.dat.
        VH: Interfaced MadLoop to Samurai and Ninja (the latter is now the default)
        HS: Turn IREGI to off by default
        MZ: new NLO generation mode. It is more efficient from the memory and CPU point of 
            view, in particular for high-multiplicity processes. 
            Many thanks to Josh Bendavid for his fundamental contribution for this.
            The mode can be enabled with
            > set low_mem_multicore_nlo_generation True
            before generating the process.
        OM: Adding the possibility to use new syntax for tree-level processes:
            QED==2 and QCD>2: the first allows to select exactly a power of the coupling (at amplitude level),
            while the second asks for a minimum value.
        RF: In the PDF uncertainty for fixed-order NLO runs, variations of alphaS were not included.
        OM: In MLM matching, fix a bug where the alpha_s reweighting was not fully applied on some events. 
            (This was leading to effects smaller than the theoretical uncertainty)
        OM: Fixing the problem of using lhapdf6 on Mac
        MZ: Faster interface for LHAPDF6
        OM: Add support of epsilon_ijk in MadSpin
        OM: Fix multiple problems with multiparticles in MadSpin
        OM: Improve spinmode=None in MadSpin
        OM: Update the TopEffTh model
        MZ: Fix problem with slurm cluster
        OM: Improve scan functionalities
        PT: New way of handling Pythia8 decays
        RF: Fixed a bug that resulted in wrong event weights for NLO processes when requiring
            a very small number of events (introduced in 2.3.3)
        OM: Allow to keep the reweight information in the final lhe file for future computation 
        MZ: updated FJcore to version 3.1.3 (was 3.0.5)

2.3.3 (15/10/15)
        OM: Allow new syntax for the param_card: instead of an entry you can enter scan:[val1, val2,...]
            To perform a scan on this parameter.
        OM: Having two modes for "output pythia8" one (default) for pythia8.2 and one for pythia8.1 (with --version=8.1)
        RF: Rewriting of job-control for NLO processes. Better accuracy estimates for FO processes
        RF: Fix for factorisation scale setting in FxFx merging when very large difference in scale in the
            non-QCD part of a process. 
        RF: Better discarding of numerical instabilities in the real-emission matrix elements. Only of interest for
	    processes which have jets at Born level, but do not require generation cuts (like t-channel single-top).
        RF: Added an option to the run_card to allow for easier variation of the shower starting scale (NLO only).
        RF: Fixed a problem in the setting of the flavour map used for runs with iAPPL >= 1. 
        RF: Allow for decay processes to compute (partial) decay widths at NLO accuracy (fixed order only).
        OM: (SysCalc interface) Allow to bypass the pdf reweighting/alpsfact reweighting
        MZ: fixed bug related to slurm clusters
	OM: remove the addmasses.py script from running by default in gridpack mode.
            if you want to have it running, you just have to rename the file madevent/bin/internal/addmasses_optional.py to
            madevent/bin/internal/addmasses.py and it will work as before. (Do not work with SysCalc tag)
        OM: make the code compatible with "python -tt" option

2.3.2.2 (06/09/15)
        VH: Finalized the MG5aMC-GoSam interface

2.3.2(20/08/15)   
        OM: Improve reweighting module. (https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Reweight)
            New Capabilities: 
             - allow to reweight by loop-induced processes
             - allow to change model 
             - allow to change process definition
             - allow to unweight the sample to have an idea of the statistical power.
             - allow to perform some crude reweighting on NLO sample (reweighting by LO matrix element). NLO
               accuracy is not preserved (in general) for such computation.
            New dependencies:
	     - require the f2py module (part of numpy)
	OM: change the kt-durham cut (at LO) such that particles coming from decays are not impacted if cut_decays
            is on False.
        VH: Fixed the check on the helas wavefunction appearance order in a helas diagram. It failed in cases
	    where additional wf were created during the fix of the fermion flow in the presence of Majorana fermions.
	RF: Fixed a bug in the aMCFast/ApplGrid interface introduced in the previous version.
        OM: Fix a crash when using mssm-no_b_mass model (due to the SLHA1-SLHA2 conversion)
        OM: Fix a bug in the add_time_of_flight function (not called by default) where the displaced vertex information
            was written in second and not in mm as it should. Note that this function can now be run on the flight
	    by adding the following line in the run_card: "  1e-2 = time_of_flight #threshold for the displaced vertex" 
	RF: Small fix that leads to an improvement in the phase-space generation for NLO processes
        OM: Fix a crash introduced in 2.3.0 when running sequentially in the same directory (thanks Gauthier)
        OM: Improve aloha in the case where some expression reduces to a pure float.
        OM: In MadSpin, allow to specify cuts for the 1>N decay in spinmode=none.
	RF: Fixed a bug that gave bogus results for NLO runs when using an internal PDF which is not
            NNPDF (like for the old cteq_6m, etc).
	RF: Fixed a bug in the PDF combination in the HwU histograms: there was no consistent use of Hessian
            and Gaussian approaches for MSTW/CTEQ and NNPDF, respectively.
        OM: Fixed a small bug in EWdim6 which was removing a coupling in the AZHH interaction.
        OM: improve customize_model function to avoid problem with unity coupling.
	RF: Improved the treatment of the bottom Yukawa. Thanks Marius Wiesemann. 

2.3.1  
     OM+VH: Automation of event generation for loop-induced processes.
		OM: Automatic change of the p/j definition to include the b particle if the model has a massless b.
	RF: Reduce the collision energy for the soft and collinear tests: for a 100TeV collider many were failing
	    due to numerical instabilities. 
        OM: Fixing a bug associated to the epsilon_ijk structure
        OM+VH: Various bug fixing for the loop-induced processes
        OM: Fix a crash in MadWidth which occurs for some 4 body decays
        PT: Fixed a bug concerning the use of Herwig++ with LHAPDF. Bug was introduced in 2.3.0.beta
	OM: Fix a crash in ALOHA for form-factor in the presence of fermion flow violation

2.3.0.beta(10/04/15) OM+VH: Adding the possibility to compute cross-section/generate events for loop-induced process
		JB+OM: Adding matchbox output for matching in the Matchbox framework
                OM+VH: Change the handling of the run_card.
                      - The default value now depends on the process you are running
                      - cut_decays is now on False by default
                      - nhel can only take 0/1 value. 1 is a real MC over helicity (with importance sampling)
                      - use_syst is set on by default (except for matching where it is kept off)
                      - New options added: dynamical_scale_choice, it can take the following value
		            -1 : MadGraph5_aMC@NLO default (different for LO/NLO/ ickkw mode) same as previous version. 
                             0 : Tag reserved for a user-defined dynamical scale (needs to be added in setscales.f).
                             1 : Total transverse energy of the event.
                             2 : sum of the transverse mass 
                             3 : sum of the transverse mass divided by 2
			     4 : \sqrt(s), partonic energy 
                OM: Cuts are also applied for 1>N processes (but the default run_card doesn't have any cuts).
                PT: Set command available for shower_card parameters
                OM: New MultiCore class with better thread support
                RF: Fixed a bug in the aMCfast/APPLGrid interface introduced in version 2.2.3
		RF: Fixed a bug in the setting of the integration grids (LO process generation) for the minimum
		    energy needed for photons. The bug did not lead to any bias in event generation.
		RF: Re-factoring of the structure of the code for NLO+PS computations.
		RF+VH: Replaced the default topdrawer histograms with HwU histograms for f(N)LO runs
                    and allow it also for aMC@NLO runs.
		RF+VB: Allow for variable bin-sizes in MG5_aMC+aMCfast+ApplGrid runs.
                MZ+RF: Added 'LOonly' asNLO mode to export processes without any real and virtuals 
                       (useful e.g. for higher multiplicities when merging)
		RF: Added support for the computation of NLO+NNLL jet veto cross sections
		RF: Fixed a bug in the Pythia8 interface: FxFx was not correctly initialized and all
                    events were incorrectly kept (introduced in v.2.2.3)
                OM: Improve the function "print_result" (in the running interface)
                    add an option --format=short allowing to print the result in a multi-column format
                OM: Possibility to not transfer pdf file to the node for each job. 
                       This is done via a new option (cluster_local_path) which should contain the pdf set.
                       This path is intended to point to a node specific filesystem.
                    New way to submit jobs on a cluster without writing the command file to disk.
                OM: Allowing MadSpin to have a mode without full spin-correlation but handling three (and more) 
                    body decay. (set spinmode=none).
                OM+PA: Fixing various bugs in MadSpin.

2.2.3(10/02/15) RF: Re-factoring of the structure of the code for fNLO computations.
                OM: Fix a bug in MadWeight (correlated param_card was not creating the correct input file)
		RF: When requiring more than 1M events for (N)LO+PS runs, do not go to higher precision than 0.001
                    for the grids and cross section (can be overwritten with the req_acc run_card parameter).
		RF: Make sure that reweight info (for PDF and scale uncertainties) also works for UNLOPS events.
		RF: When setting the B's stable in the shower_card, also set the eta_b (PDG=551) stable.
		OM: Change the Breit-Wigner splitting for the multi-channel integration, use the bwcutoff instead of
                    the hardcoded value 5.
                MZ: Fix to bug 1406000 (segfault appearing when doing FxFx merging). Thanks to Josh Bendavid for
                    having reported it
                MZ: Fix to a bug occurring when generating event in the "split" mode: the required output was 
                    not correctly specified
                OM: The built-in pdfs "nn23lo" and "nn23lo1" were associated to the wrong lhapdfid in the lhef file.
                    This was creating a bias when using SysCalc. (Thanks Alexis)
                OM: Fix a bug in the LO re-weighting module which was removing the
                    SysCalc weight from the lhe file (thanks Shin-Shan)
                Team: Fixes to different small bugs / improvements in the error and warning messages
		RF: For aMC runs, If a NAN is found, the code now skips that PS point and continues instead of
		    leading to NAN.
                RF: For fNLO runs the virtuals were included twice in the setting of the integration grids. 
                    This was not leading to any bias in previous version of the code.

2.2.2(06/11/14) OM: Correct a bug in the integration grid (introduced in 2.1.2). This was biasing the cross-section of
                    processes like a a > mu+ mu- in the Effective Photon Approximation by three orders of magnitude.
                    For LHC processes no sizeable effect has been observed so far.
                MZ: some information for aMC@NLO runs which was before passed via include files is
                    now read at runtime. The size of executables as well as compilation time / memory usage
                    is reduced for complicated processes
                RF: Fix crash #1377187 (check that cuts were consistent with the grouping was too restrictive) 
		RF: For NLO running: added 'strip' to the makefiles to reduce executable sizes (removes symbol info)
		Stefano Carrazza (by RF): fix for the photon PDF for the internal NNPDF sets
		RF: Improved the check on the consistency of the cuts and the grouping of subprocesses (LO running)
                PT: enabled PYTHIA8.2
                OM: restore the usage of the external gzip library for files larger than 4Gb which were crashing with
                    the python gzip library
                OM: Fixing the default card for Delphes
                OM: Improve support of the lsf cluster (thanks Josh)
                OM: Adding support for the UFO file functions.py (which was ignored before)
                OM: Reduce the amount of RAM used by MadSpin in gridpack mode.
                OM: discard in MadWidth partial widths lower than \Lambda_QCD for colored particles.

2.2.1(25/09/14) OM: Fix a bug preventing the generation of events at LO due to a wrong treatment of 
                      the color-flow.

2.2.0(24/09/14) VH: General mixed order corrections in MadLoop (only), including QCD/EW/QED and 
                    the UFO@NLO model 'loop_qcd_qed_sm'.
                VH: Re-design of both the tree and MadLoop matrix elements output to compute
                    contributions of different coupling order combinations independently from one another.
                VH+HS: Tensor integral reduction as implemented in PJFry and IREGI readily available
                    and interfaced to MadLoop's output.
                VH: Re-structuring of MadLoop's standalone output so as to easily create a single dynamic 
                    library including many processes at once. Useful for interfacing MadLoop to other MC's
                    and already working with Sherpa.
                VH+HS: This branch contains all the fixes for proper treatment of the latest BSM@NLO models
                    produced by FeynRules@NLO. In particular, the fixes related to the presence of Majorana
                    particles in loop ME's.
                RF: Corrected the behaviour of the pdfcode parameter in the shower_card for NLO+PS runs.
                PT: Redesigned shower_card.dat and eliminated modbos options for Herwig6          
                RF: Change the SPINUP information in the NLO LHEF from 0 to 9 (i.e. sum over helicities)
                RF: Fixed a bug in the check on the determination of the conflicting BWs.
		RF: Added the aMCfast+APPLgrid interface (arXiv:1406.7693 [hep-ph])
                PT: Redesigned shower_card.dat and eliminated modbos options for Herwig6          
                RF: Change the SPINUP information in the NLO LHEF from 0 to 9 (i.e. sum over helicities)
                RF: Fixed a bug in the check on the determination of the conflicting BWs.
                MZ: enabled LHAPDF6 interface 
                OM: Fixed a crash in some HEFT merging case.
                OM: Fix various compatibility problems created by the LHEFv3 version (Thanks to S. Brochet)
                OM: Fix a bug for MadSpin in gridpack mode
                OM: Add a routine to check the validity of LHE file (check_event command)
                OM: Fix a bug for UFO models with custom propagators
                OM: Fix a bug in the computation of the cross-section in the presence of negative contributions
                OM: Change colorflow information of LHE file in the presence of two epsilon_ijk
                    since PY8 was not able to handle such flow in that format.
                OM: Add the function print_result for aMC@(n)LO run.
                OM: Add some shortcut in the card edition 
                    set ebeam 500 # edit both beams
                    set lpp 0     # edit both lpp1 and lpp2
                    set lhc 14    # configure for LHC 14TeV
                    set ilc 1000  # configure for ilc 1TeV
                    set fixed_scale 100 # set all scale to fixed and at 100GeV
		    set showerkt T # set showerkt on T in the shower card
 		    set qcut 20    # set the qcut to 20 in the shower card
                OM: Fix a bug in the card edition mode which was sometimes returning to default values
                    which had been edited by hand and not via the set command.
                Seoyoung Kim (by OM): Implementation of the htcaas (super-)cluster support.
		Juan Rojo (by RF): extended the 3 internal NNPDF sets for scales relevant for a 100TeV collider.
                OM: Fix a problem with the creation of DJR plot with root 6
                OM: allow to set the width to Auto in NLO computations (width computed at LO accuracy)
                OM: Adding the possibility to have automatic plot after the parton shower for Herwig6/Pythia6.
                    This requires MadAnalysis and the pythia-pgs package.

2.1.2(03/07/14) OM: Fix a bug in ALOHA in the presence of customized propagator (Thanks Saurabh)
                OM: Fixing some compilation issues with MadWeight (Thanks A. Pin)
                OM: Fixing a bug preventing MadWidth to run due to the model prefixing (depending
                    on the way it was called)
                OM: Fixing a bug in MadSpin in the mssm model
		RF: Added the invariant mass and DeltaR cuts for *same flavour* opposite sign lepton
                    pairs to the run_card for NLO-type generations.
		RF: Added FxFx and UNLOPS merging with Pythia8
		RF: Prevent an infinite loop in MadSpin by forcing the correct sign to the invariants
		RF: Catch a possible error related to grouping subprocesses and setcuts
		OM: Fix an error when using the "customize_model" command
                S. Mrenna (by OM): Fix the include file in pythia8 output to be compliant with the latest
                    PY8 version
		RF: Added a string with functional form for the scales to the event file banner (NLO only)
                S. Brochet (by OM): Fix a bug in MadSpin with the writing of the mother ID in the LHE file.
                    Force the tag in the banner to always have the same case
                    increase momenta precision for the LHE file written by MadSpin 
                    (thanks a lot to S. Brochet for all those patches)
                PT: Integrated Jimmy's underlying event for Herwig6
                OM: improve "add model" functionality allow to force particle identification.
                PT: Bug fix in the normalisation of topdrawer plots for option 'sum' (as opposed to 'average')
		RF: Fixed a bug related to the random seed when the code was not recompiled for a new run.
                OM: Fixed a bug in MadEvent(LO) runs: the generated samples were biased in the presence of
                    negative cross-sections. A negative cross-section is possible only if you use a NLO PDF 
                    and/or if you edit the matrix.f by hand to have a non-definite positive matrix-element.
		OM: When importing a model, check that there is not more than 1 parameter with the same name.
                PT: Substantial recoding of montecarlocounter.f and of a subroutine in fks_singular.f. Will help 
                    future extensions like EW NLO+PS matching and numerical derivatives      
                OM: Fixing a wrong assignment in the color flow in the presence of the epsilon_ijk color structure.
                    Those events were rejected by Pythia8 due to this wrong color-flow.
                MZ: Added the possibility to run the shower on a cluster, possibly splitting the lhe file 
                MZ: The c++ compiler can be specified as an option in the interface. On MACOSX, clang should
                    work now
                OM: MadEvent output is now LHEFv3 fully compliant. A parameter in the run_tag (lhe_version) 
                    allows to return LHEF version 2 format for retro-compatibility.

2.1.1(31/03/14) OM: Change the way the UFO model is handled by adding a prefix (mdl_) to all model variables.
                    This avoids any potential name conflict with other parts of the code. This feature can be
                    bypassed by using the option --noprefix when importing the model.
                OM: New command "add model XXX" supported. This command creates a new UFO model from two UFO model.
                    The main interest stands in the command "add model hgg_plugin", which adds the effective operator
                    h g g to the original UFO model. The model is written on disk for edition/future reference.
		RF: Reduced the calls to fastjet and skipped the computation of the reweight coefficients when
 		    they are not needed.
                OM: Fixed a bug for LO processes where the MMLL cut was not applied to the event sample.
                PA: Fix a bug in MadSpin to avoid numerical instabilities when extracting t-channel invariants
                    from the production event file (see modification in driver.f, search for 'MODIF March 5, 2014') 
                OM: Better determination of which particles are in representation 3/3bar since FR is ambiguous on that point.
                    Now the determination also looks for 3 -3 1 interactions to check if that helps.
                OM: Fix a bug(crash) in MW linked to the permutation pre-selection module.
		RF: Better comments in the code for user-defined cuts in the ./SubProcesses/cuts.f function.
                    Also the maxjetflavor parameter in the run_card is now actually working.
                OM: Update SysCalc to:
                      - Fix a bug where some files were sometimes truncated.
                      - Allow for independent scale variation for the factorization/renormalization scale.
                RF+OM: Improve the handling of conflicting Breit-Wigners at NLO
		RF: Print the scale and PDF uncertainties for fNLO runs in the summary at the end of the run

2.1.0(21/02/14) MADWEIGHT RELEASE:
                ------------------
                
                OM+PA: First Official release of MadWeight inside MG5_aMC
                      Main update:
                        - ISR corrections
                        - possibility to use narrow-width approximation
                        - introducing a module for the pre-selection of the parton-jet assignment.
                        - extended formalism for the transfer function (more generic)
                        - possibility to evaluate the weights for multiple choices of transfer function 
      			  on the same phase-space point. The phase-space is optimized for the first set of 
                          parameters.
		      Speed update:
                        - More efficient way to group the computation for identical processes with different final states.
                        - Possibility to Monte-Carlo over the permutation.
                        - More efficient way to choose between the various change of variable.
                        - Possibility to use mint (not compatible with all of the options)
			- Possibility to use sobol for the generation of PS points (sometimes faster than a pure
                          random point generator).

                MadEvent/aMC@NLO UPDATE/BUG FIXING:
 		-----------------------------------

                OM: Fix critical bug (returns wrong cross-section/width) for processes where the center of mass 
                    energy of the beam is lower than 1 GeV. So this has no impact for LHC-collider phenomenology.
                    This can also impact the computation of decay widths if the mass of that particle is below 1 GeV.
		RF: Critical bug fixed (introduced in 2.0.2) for fixed order NLO runs that could
		    give the wrong cross section when the phase-space generation is inefficient
                    (like in the case for conflicting Breit-Wigners). This bug did not affect runs
                    with matching to the parton shower.
                OM: Fix a bug leading to a crash with some decay syntax. i.e., p p > t t~, (t > w+ b), (t~ >w- b~)
                OM: Fix format of LHE output for 1>N events when the <init> and mother information were wrongly set 
                    to LHC default. Specific support of this option will be part of pythia8 (8.185 and later)
                OM: Fix the syntax for the custom propagator to follow the description of arXiv:1308.1668 
                OM: Allow to call ASperGe on the fly if the ASperGe module is included in the UFO model.
                    just type "asperge" at the moment when the code proposes that you edit the param_card.

                MADSPIN UPDATE:
                ---------------
                OM: Allow to use another model for the decay than the one used for the production of events.
                    You are responsible for the consistency of the model in that case.
                PA: Include helicity information for the events generated by MadSpin.
                OM: Fix a bug in MadSpin preventing the gridpack to run with NLO processes.

2.0.2(07/02/14) RF: Suppressed the writing of the 'ERROR in OneLOop dilog2_r' messages (introduced in the 
                    previous version)
                OM: Fix the bug that the shower_card.dat was wrongly identified as a pythia_card.
                OM: add one MadSpin option allowing to control the number of simultaneous open files.
                OM: Fix a bug in eps preventing evince from displaying labels on page 2 and following
                    Thanks to Gauthier Durieux for the fix.
                OM: Fix a bug(crash) for p p > w+ w- j j introduced in 2.0.0 due to some jets sometimes being tagged as QCD
                    and sometimes not (which was making the automatic scale computation crash)
                OM: Change the way to write the <init> line of the lhe file to take into account
                    - processes with more than 100 subprocesses (note that you need to hack the pythia-pgs
                      package to deal with such a large number of sub-processes)
                    - deal with pdf identification number bigger than 1 million.  
                OM: Fixed a bug preventing MadEvent from detecting external modules (pythia-pgs, syscalc,...)
                    Bug #1271216 (thanks Iwamoto)
                PT: PYTHIA8 scale and pdf variations

2.0.1(20/01/14) OM: Fix a bug in h > l+ l- l+ l- for group_subprocesses=False (decay only). A follow up of 
                    the bug fix in 2.0.0
                RF: Replaced the Error#10 in the generation of the phase-space (for NLO) to a Warning#10.
                    In rare cases this error stopped the code, while this was not needed.
                RF: When using non-optimized loop output, the code now also works fine.
                OM: Modification of the code to allow it to run on our servers
                VH: Improve the timing routine of the NLO code (displayed in debug mode)
                VH: FIX the import of old UFO models (those without the all_orders attribute).
                OM: Add a functionality for restrict_model: if a file paramcard_RESTRICTNAME.dat
                    exists, then this file is used as the default param_card for that restriction.
                HS: Updated CutTools to v1.9.2

2.0.0(14/12/13)    CHANGE IN DEFAULT:
                   ------------------
                      OM: Change the Higgs mass to 125 GeV for most of the models (except the susy/v4 ones).
                      OM: Change the default energy of collision to 13 TeV.
                      RF: Default renormalisation and factorisation scales are now set to H_T/2. (for aMC only)

                   MadEvent Update:
                   ----------------
                      OM+SF+RF: Add Frixione Photon isolation (also for aMC)
                      OM: Implementation of the reweight module for Leading Order matrix-element
                      JA+OM+AK: Store parameters needed for systematics studies.
                          This can be turned on with the use_syst parameter in
                          run_card.dat.
                          This output can be used to generate event weights for
                          a variety of variational parameters, including scalefact,
                          alpsfact, PDF choice, and matching scale. Note that this requires
                          pythia-pgs v2.2 for matching scale.
                      OM+JA+Chia: Implement MadWidth (automatic/smart computation of the widths)
                      OM: Support for Form-Factors defined in the UFO model, and support for the presence of model
                          parameters inside the Lorentz expression.
                      OM: Support for an arbitrary functions.f file present inside the UFO model.
                      JA: Included <clustering> tag in matched .lhe output, to be 
                          used together with Pythia 8 CKKW-L matching. This can be 
                          turned off with the clusinfo flag in run_card.dat.
                      JA: New treatment of matching for diagrams that have no
                          corresponding lower-multiplicity diagrams. Jets that
                          are not classified as shower-type emission jets are
                          flagged in the cluster scale info at the end of the event,
                          which is recognized by the Pythia interface in Pythia-PGS
                          package v. 2.2. For such jets, xqcut does not apply. This
                          allows for consistent matching e.g. of p p > w+ b b~ in 
                          the 4-flavor scheme. Note that auto_ptj_mjj must be set to
                          .false. for this to work properly.
                      OM: Change model restriction behavior: two widths with identical values are not merged anymore.
                      S.Prestel(via OM): implement KT Durham cut. (thanks to Z. Marshall)
                      OM: Improved check for an unresponsive PBS cluster (thanks J. Mc Fayden)
                      OM: Implement a maximum number (2500) of jobs which can be submitted at the same time
                          by the PBS cluster. This number is currently not editable via configuration file.
                     
                   MadEvent Bug Fixing:
                   --------------------
                      OM: Fix a bug for h > l+ l- l+ l- (introduced in 1.5.9) where the phase-space parametrization
                          fails to cover the full phase-space. This bug occurs only if two identical particles decay
                          into identical particles and if both of those particles can't be on-shell simultaneously.
                      OM: Fix a bug for multi_run sample in the presence of negative weights (possible if NLO pdf)
                          The negative weights were not propagated to the merged sample. 
                          (thanks to Sebastien Brochet for the fix)
	         
                   aMC@NLO Update:       ! FIRST OFFICIAL RELEASE WITH NLO CAPABILITIES !
                   ---------------
                       PT: MC@NLO matching to PYTHIA8 available.
                       RF: Added FxFx merging
                       RF: Use MC over helicities for the MadLoop virtual corrections.
                       RF: Using "virtual tricks" to reduce the number of PS points for which to include
                           the virtual corrections, leading to a speed up of the code.
                       OM+SF+RF: Add Frixione Photon isolation (also in MadEvent)
                       PA+OM: Fast version of MadSpin implemented (PS generation in Fortran).
		       OM: Allow to have MadSpin in "gridpack mode" (same cards/same decay). 
                           Add in the madspin_card "set ms_dir PATH". If the path didn't exist MS will
                           create the gridpack on that path, otherwise it will reuse the information 
                           (diagram generated, maximum weight of each channel, branching ratio,...)
                           This allows to bypass all the initialization steps BUT is valid only for the
                           exact same event generation.
                       VH: Fixed set_run.f which incorrectly sets a default value for ptl, drll and
                           etal making the code insensitive to the values set in the run_card.dat 
                       VH: Fixed a bug in MadLoop that doubled the computing time for quad precision
                       VH+RF: Added MadLoop stability information to the log files (and run summary
                           in DEBUG mode).
                       RF: Fixed a stability problem in the reweighting to get PDF and scale uncertainties.
                       VH+RF: Improved the Binoth LHA interface
                       RF: Improved the multi-channeling for processes with more amplitudes than diagrams.
                       RF: Added a new parameter in the run_card to set the required accuracy for fixed 
                           order runs.
                       SF+RF: Improved handling of fixed order analysis

                    From beta3 (13/02/13):
                       OM: Merge with 1.5.7+1.5.8 (See information below)
                       OM: Allow the customize_model to be scriptable in a 
                           friendly way.
                       RF: Event normalization is now Les Houches compliant (the weights
		           of the events should average to the total rate). The old normalization
                           can still be chosen by setting the flag 'sum = event_norm' in the run_card.
		       RF: Fixes a bug related to the mass of the tau that was not consistently 
 		           taken into account in the phase-space set-up.
		       VH: Fixed the incorrect implementation of the four gluons R2 in the loop_sm UFO.
		       VH: Fixed the UV renormalization for the SM with massive c quarks.
                       RF: The PDF uncertainty for NNPDF is now also correctly given in the run summary
                       RF: Some improvements in the test_MC and test_ME when matrix elements are
                           numerically very large
                       OM+RF: Added the running at LO to the 'launch questions'
                       OM: Allow "check" command to use a event file.
                           This will use the related param_card and the first
                           event compatible with the requested process.
                       RF: Improved the phase-space generation in the case of competing resonances

                    From beta2 (23/12/12):
                       MG5 Team: Include 1.5.4+1.5.5+1.5.6 modifications
                       MadSpin Team: Include MadSpin
                       VH: Fix computation in the Feynman gauge for the loops
                       RF: automatic computation of the NLO uncertainties
                       OM: NLO can now be run with no central disk
                       MZ: change the format of number (using e and not d)
                       MZ: compilation and tests are possible in multicore
                       RF: allow to specify either the uncertainty or the number of events
                           for aMC@NLO/NLO
                       OM: ./bin/mg5 cmd.cmd is now working for NLO process

                    From beta1 (31/10/12):
                       aMCatNLO Team: First public (beta) version of aMCatNLO.
                         In order to learn aMCatNLO, please do "tutorial aMCatNLO"
                         Please also visit: http://amcatnlo.cern.ch/list.htm for more
                         information.

1.5.15 (11/12/13) OM: Fix the auto-update function in order to allow to pass to 2.0.0

1.5.14 (27/11/13) OM: Add a warning about the fact that newprocess_mg5 is going to be removed in MG5_aMC_V2.0.0
                  OM: Improved cluster submission/re-submission control.

1.5.13 (04/11/13) OM: Implement a function which checks if jobs submitted to a cluster ran correctly.
                      In case of failure, you can re-submit the failing jobs automatically. The maximal
                      number of re-submissions for a job can be parametrized (default 1) and how long you have to
                      wait before this resubmission [to avoid slow filesystem problem, i.e. condor](default 300s)
                      Supported cluster for this function: condor, lsf, pbs
                  OM: Fix a problem when more than 10k diagrams are present for a given subprocess.
                      (tt~+4jets).
                  BF: Change the nmssm model (the coupling orders were not correctly assigned for some triple
                      Higgs interactions) 
                  OM: use evince by default to open eps file instead of gv.
		  OM: Fix a problem with the set command for the card edition for the mssm model.
                  OM: Update the EWdim6 model according to the snowmass paper. (3 more operators)
                      The default model is restricted in order to exclude those operators. In order
                      to have those you have to use import model EWdim6-full
                  OM: Fix bug #1243189, impossible to load v4 model if a local directory has the name of
                      the models (which is present in the models directory)
                  OM: Fix a bug in the complex mass scheme in the reading of the param_card (it was clearly stated)
                  OM: Improve numerical stability of the phase-space point generation. (thanks Z. Surujon)

1.5.12 (21/08/13) OM: Improve phase-space integration for processes with a strong MMJJ cut. Cases where
                      the cross-section was slightly (~4%) under-evaluated due to such a strong cut.
                  OM: Add a command print_results in the madevent interface. This command prints the
                      cross-section/number of events/... 
                  OM: change the way the prompt color is handled (no systematic reset), which provides better
                      results when the log is printed to a file. (thanks Bae Taegil) 
                  OM: Fix Bug #1199514: Wrong assignment of mass in the lhe events file if the initial 
                      state has one massive and one massless particle. (Thanks Wojciech Kotlarski)
                  OM: Fix a compilation problem for SLC6 for the installation of pythia-pgs
                  OM: Fix a crash linked to bug #1209113.
                  OM: Fix a crash if python is not a valid executable (Bug #1211777)
		  OM: Fix a bug in the edition of the run_card if some parameters were missing in the cards
                      (Bug #1183334)

1.5.11 (21/06/13) OM: Fix CRITICAL bug (returning wrong cross-section) for processes with more than
                      one W decaying leptonically. For such processes the lepton cuts were also used
                      on the neutrino particle reducing the cross-section. This bug was present only
                      for group_subprocesses=True (the default)
                  OM: Fix Bug #1184213: crash in the presence of GIM mechanism (occur on some 
                      LINUX computer only)
                  OM: The compilation of madevent is now performed using the number of cores specified
                      in the configuration file. Same for pythia, ...
                  OM: Improve support for Read-Only system
                  OM: Fix a bug with the detection of the compiler when the user specifies a specific
                      compiler.
                  OM: Fix a problem where MG5 fails to compute the cross-section/width after
                      a first computation fails to integrate due to a wrong mass spectrum.
                  OM: Fix a wrong output (impossible to compile) for pythia in the presence of photon/gluon
                      propagators (introduced in 1.5.8)
                  OM: Allow to have UFO model with "goldstone" attribute instead of "GoldstoneBoson", since
                      FR changed convention in order to match the UFO paper.

1.5.10 (16/05/13) OM: Fix Bug #1170417: fix a crash for the conjugate routine in the presence of
                      massless propagators. (introduced in 1.5.9)
                  OM: Fix question #226810: checking that patch program exists before
                      trying to update MG5 code.
                  OM: Fix Bug #1171049: an error in the order of wavefunctions
                      making the code crash (introduced in 1.5.7)
		  OM: Allow to use an additional syntax for the set command.
                      set gauge = Feynman is now valid. (Was not valid before due to the '=')
                  OM: Fix by Arian Abrahantes. Fix the SGE cluster which was not working when
                      running full simulation (PGS/Delphes).
                  OM: adding txxxxx.cc (Thanks to Aurelijus Rinkevicius for having 
                      written the routine) 
                  OM: Fix Bug #1177442. This crash occurs only for very large models.
                      None of the models shipped with MG5 are impacted.
                  OM: Fix Question #228315. On some filesystems, some of the executables
                      lose the permission to be executable. Recover those errors 
                      automatically.
                  OM: Modify the diagram enhancement technique. When more diagrams have
                      the same propagator structure we still combine them but we now include
                      the interference term in the enhancement technique for those diagrams.
                      This fixes a crash for some multi-jet processes in the presence of non-diagonal
                      ckm matrices.

1.5.9 (01/04/13)  JA: Fix bug in identification of symmetric diagrams, which could
                      give the wrong propagators included in event files for certain
                      processes (such as p p > z z j, z > l+ l-). Apart from the 
                      propagators (with status 2) in the event file, this bug didn't
                      affect any other results (such as distributions).
                  JA: Fix problem in gensym which made some decay chain processes
                      slower than they should be. Thanks Eric Mintun for reporting.
                  JA: Fix problem in event clustering (introduced in v. 1.5.8)
                      which made events from some processes fail Pythia running.
                  JA: Fixed bug #1156474, Pythia 8 C++ matrix element output for 
                      decay chain processes such as p p > z j, z > j j.
                      (Bug #1156474)
                  JA + OM: Automatically remove libpdf and libgeneric before survey,
                      so everything works automatically when switching between
                      built-in PDFs and LHAPDF.
                  JA: Allow syntax / to remove particles in the define command.
                      Example: define q = p / g
                  JA: Added fat warning if any decay process in a decay chain
                      includes a particle decaying to itself (as is the case
                      if you do w+ > all all, since you include w+ > w+ a).
                  JA: Forbid running newprocess_mg5 from a process directory
                      that has already been generated, to avoid confusion.
                  OM: Fix lxplus server issue (Bug #1159929)
                  OM: Fix an issue when MG5 directory is on a read only disk 
                      (Bug #1160629)
                  OM: Fix a bug which prevented having the pythia matching plot/cross-section
                      in some particular case.
                  OM: Support of new UFO convention allowing to define custom propagator.
                      (Both in MG5 and ALOHA)
                  OM: Change ALOHA default propagator to have a specific expression for the
                      massless case allowing to speed up matrix element computation with 
                      photon/gluon.
                  OM: Correct the default spin 3/2 propagator (wrong incoming/outgoing
                      definition)
                  ML (by OM): Adding support of the SLURM cluster. Thanks to 
                      Matthew Low for the implementation.
                  OM: Fixing the standalone_cpp output for the mssm model. (only model impacted)
                      Thanks to Silvan S Kuttimalai for reporting. 
                  OM: Fix Bug #1162512: wrong line splitting in cpp when some names were very long.
                      (shorten the name + fix the splitting)

1.5.8 (05/03/13)  OM: Fix a critical bug introduced in 1.5.0. ALOHA was wrongly writing
                      HELAS routines for expressions containing a squared expression
                      (like P(-1,1)**2). None of the default models of MG5 (like sm/mssm)
                      have such a type of expression. More information in bug report #1132996
                      (Thanks Gezim) 		
                  OM+JA: install Delphes now installs Delphes 3 
                      [added command install Delphes2 to install Delphes 2]
                  OM: Add command in MadEvent interface: add_time_of_flight
                      This command modifies the lhe events file by adding the time of 
                      flight information in the lhe events. To run this you need to do
                      $> ./bin/madevent
                      MGME> generate_events --laststep=parton -f 
                      MGME> add_time_of_flight
		      MGME> pythia    [if needed]
                  OM: Fix bug in pythia8 output for process using decay chains syntax.
                      See bug #1099790.
                  CDe+OM: Update EWdim6 model
                  OM: Fix a bug preventing models customized via the "customize_model"
                      command from using the automatic width computation.
                  OM: Change model restriction behavior: a value of 1 for a width is 
                      not treated as a restriction rule.
                  OM: Fix incomplete restriction of the MSSM model leading to inefficient
                      process merging (and larger-than-necessary files) for the MSSM.
                  OM: Correct bug #1107603 (problem with condor cluster for submission 
                      associated to a large number of jobs). Thanks Sanjay.
                  JA: Fix one part of the problem in bug #1123974: take into 
                      account invariant mass cuts mmXX above the peak range in 
                      s-channel resonances in the phase space integration,
                      to make sure such channels find events even for narrow
                      invariant mass cut ranges. Please note the discussion in 
                      that bug report for other types of channels however.
                  JA: Fixed bug #1139303: matrix elements for identical 
                      decay chain processes with different propagators (such as 
                      p p > w+, w+ > e+ ve and p p > w-,  w- > vm~ mu-) 
                      are now no longer combined, to ensure that resonances are
                      correctly represented in event file.
                  OM: Support lhapdf set which contains photon (See bug #1131995).
                  RF+JA: Reuse last two PDF calls also for LHAPDF PDFs, clarify code
                      for reuse of PDFs in pdg2pdf.f and pdg2pdf_lhapdf.f
                  OM: Update the default delphes card to latest Delphes version. This 
                      default card is automatically overwritten by the default Delphes
                      card when running "install Delphes".
                  JA: Make sure cuts are only checked once per event - this can save
                      a lot of time for multiparton event generation.
                  OM: Fix Bug #1142042 (crash in gridpack).

1.5.7 (15/01/13)  OM+JA: Fixed a crash linked to model_v4 for processes containing wwww or
                      zzww interactions. (See bug #1095603. Thanks to Tim Lu) 
                  OM: Fix a bug affecting 2>1 processes when the final state particle is
                      an outgoing fermion, introduced in version 1.5.0. (Thanks to 
                      B. Fuks) 
                  OM: Fix a problem of fermion flow for v4 model (thanks to A. Abrahantes) 
                  OM+DBF: Change automatically the electroweak scheme when passing to the
                      complex-mass scheme: the mass of the W is an external parameter
                      and Gf is an internal parameter fixed by LO gauge relation. 
                  OM+DBF: Remove the model sm_mw from the model database.
                  OM: Fix a problem in the ./bin/mg5 file command when some questions are
                      present in the file.
                  OM: Extend support for ~ and ${vars} in path.
                  OM: Fix a crash in multi_run for more than 300 successive runs.
                      (Thanks to Diptimoy)
                  OM: Allow to choose the center of mass energy for the check command.
                  OM: small change in the pbs cluster submission (see question #218824)
                  OM: Adding the possibility to check gauge/lorentz/... for 2>1 processes.

1.5.6 (20/12/12)  JA: Replaced error with warning when there are decay processes
                      without a corresponding core process final state (see
                      Question #216037). If you get this warning, please check
                      carefully the process list and diagrams to make sure you
                      have the processes you were expecting.
                  JA: Included option to set the highest flavor for alpha_s reweighting
                      (useful for 4-flavor matching with massive b quarks). Note that
                      this does not affect the choice of factorization scale.
                  JA: Fixed Bug #1089199, where decay processes with symmetric 
                      diagrams were missing a symmetry factor. 
                      Note that this only affects decay processes (A > B C ..) 
                      with multiple identical particles in the final state and 
                      some propagators not able to go on the mass shell.
                  JA: Updated the restriction cards for the sm model to set 
                      Yukawa couplings equal to the corresponding masses
                      (in order to avoid stupid gauge check failures).


1.5.5 (18/11/12)  JA: Fixed Bug #1078168, giving asymmetries in X+gamma generation
                      (e.g. Z+gamma) when ickkw=1 and pdfwgt=T. Thanks Irakli!
                  JA: Ensure that t-channel single top gives non-zero cross section
                      even if maxjetflavor=4 (note that if run with matching,
                      maxjetflavor=5 is necessary for correct PDF reweighting).
                  OM: Fixed Bug #1077877. Aloha crashing for pseudo-scalar, 3 boson
                      interactions (introduced in 1.5.4)
                  OM: Fix a bug for the command "check gauge". The test comparing
                      results between the two gauges (unitary and Feynman) was not
                      changing the gauge correctly.
                  OM: Improvement in LSF cluster support (see bug #1071765) Thanks to
                      Brian Dorney.

1.5.4 (11/11/12)  JA: Fixed bug in combine_runs.py (introduced in v. 1.5.0) for
                      processes with 5 final-state particles, which might prevent
                      matching to Pythia to work properly (thanks Priscila).
                  OM: Fix Bug #1076043, error in kinematics for e- p collisions,
 		      thanks to Uta Klein (introduced in 1.5.0).
                  JA: Fix Bug #1075525, combination of decay processes for 
                      particle and antiparticle (e.g. w+ > all all and 
                      w- > all all), thanks Pierre.
                  OM: Fix a compilation crash due to aloha (thanks Tim T)
                  JA: Fixed dynamical scale settings for e- p collisions.
                  JA: Fixed running LHAPDF on a cluster with cluster_temp_path.
                  JA: Ensure that the seed is stored in the banner even when Pythia
                      is run (this was broken in v. 1.5.0).
                  JA: Improved and clarified phase space presetting for processes
                      with competing BWs.

1.5.3 (01/11/12)  OM: Fix a crash in the gridpack mode (Thanks Baris Altunkaynak)
                  OM: Fix a crash occurring on clusters with no central disk (only
                      condor by default) for some complicated process.
                  OM: If launch command is typed before any output command, 
                      "output madevent" is run automatically.
                  OM: Fix a bug preventing setting the width to Auto in the mssm model.
                  OM: Allow "set width PID VALUE" as an additional possibility to
                      answer edit card function.
                  OM: Improve ME5_debug file (include now the content of the 
                      proc_card as well).

1.5.2 (11/10/12)  OM: Fix a bug for the mssm model. The param_card was not read properly
                      for this model. (introduced in 1.5.0)
                  OM: If the code is run with an input file (./bin/mg5 cmd.cmd)
                      All questions not answered in the file will be answered with the
                      default value. Running with piping data is not affected by this.
                      i.e. running ./bin/mg5 cmd.cmd < answer_to_question 
                       or echo 'answer_to_question' | ./bin/mg5 cmd.cmd      
                      are not affected by this change and will work as expected.
                  OM: Fixing a bug preventing the use of the "set MH 125" command in a
                      script file.
                  JA: Fixed a bug in format of results.dat file for impossible
                      configurations in processes with conflicting BWs.
                  OM: Adding command "launch" in madevent interface which is the
                      exact equivalent to the launch command in the MG5 interface
                      in madevent output.
                  OM: Secure the auto-update, since we received some reports of incomplete
                      version file information.

1.5.1 (06/10/12)  JA: Fixed symmetry factors in non-grouped MadEvent mode
                      (bug introduced in v. 1.5.0).
                  JA: Fixed phase space integration problem with multibody 
                      decay processes (thanks Kentarou for finding this!).
                  OM: Fix that the standalone output was not reading the param_card correctly
                      (introduced in 1.5.0)
                  OM: Fix a crash when trying to load heft
                  OM: Fix the case when the UFO model contains one mass which 
                      has the same name as another parameter up to the case.
                  OM: Fix a bug for results lower than 1e-100; those are now
                      considered as zero.
                  OM: Fix a bug present in the param_card created by the width
                      computation, where the qnumbers data were written as a float
                      (makes Pythia 6 crash).

1.5.0 (28/09/12)  OM: Allow MG5 to run in complex mass scheme mode
                      (mg5> set complex_mass True)
                  OM: Allow MG5 to run in Feynman gauge
                      (mg5> set gauge Feynman)
                  OM: Add a new command: 'customize_model' which allows (for a
                      selection of models) to fine tune the model to your needs.
                  FR team: add a file decays.py in the UFO format; this file
                      contains the analytical expression for one to two decays
       		  OM: implement a function for computing the 1 to 2 width on 
                      the fly. (requires MG5 installed on the computer, not only
                      the process directory)
                  OM: The question asking for the edition of the param_card/run_card
                      now accepts a command "set" to change values in those cards
                      without opening an editor. This allows simple implementation
                      of scanning. (Thanks G. Durieux for having pushed me to do it)
                  OM: Support UFO model with spin 3/2
                  OM + CDe: Support four fermion interactions. Fermion flow 
                       violation/Majorana are not yet allowed in four fermion 
                       interactions.
                  OM + PdA: Allow Large Extra Dimension Model (LED) to run in the
                      MG5 framework.
                  OM: Add auto-detection if MG5 is up-to-date and propose to
                      apply a patch if not.
                  OM: MadEvent changes automatically the compiler according to 
                      the value present in the configuration file.
                  OM: Aloha modifications: faster to create routines and more 
                      optimized routines (up to 40% faster than previous version).
                  OM: Aloha now supports Lorentz expression with denominator.
                  OM: Improve error message when Source didn't compile properly.
                  OM: The numerical evaluation of the matrix element requires now 
                      less memory than before (madevent and standalone output)
                  OM: Fix a series of bugs with the madevent command 'remove' and 
                      'run_banner'                    
                  JA: Ensure identical results for identical seeds also with
                      multiple runs in the same directory. Note that identical runs
                      with previous versions can't be guaranteed (but different
                      seeds are guaranteed to give statistically independent runs).
                      Save the results.dat files from all runs.
                  JA: Amended kinematics to correctly deal with the case of
                      massive beams, as well as fixed-target proton collisions.
                  JA: Changed default in the run_card.dat to use -1 as "no cut"
                      for the max-type cuts (etaj, ptjmax, etc.).
                  JA: Added support for negative weights in matrix elements
                      (as required for interference-only terms) and PDFs.
                  JA: Avoid creating directories for integration channels
                      that can not give events based on BW settings
                      (further improvement compared to v. 1.4.8).
                  JA: Optimize phase space integration when there are resonances
                      with mass above ECM.
                  JA: Fixed issue in replace.pl script with more than 9 particles
                      in an event.
                  JA+OM: Allow cluster run to run everything on a local (node) disk.
                      This is done fully automatically for condor cluster.
                      For the other clusters, the user should set the variable
                      "cluster_temp_path" pointing to a directory (useful only if
                      the directory is on the node filesystem). This still requires
                      access to central disk for copying, event combination,
                      running Pythia/PGS/Delphes etc.
                  OM: Replace fortran script combine_runs by a python script. 
                      This script allows to be more stable when running on clusters 
                      with slow filesystem response (bugs #1050269 and #1028844)
                  JA: Ensure that process mirroring is turned off for decay
                      processes of type A > B C...

1.4.8.4 (29/08/12) OM: Fix a web problem which caused generations to run twice on the web.

1.4.8.3 (21/08/12) JA: Ensure that the correct seed is written also in the .lhe
                       file header.
                   JA: Stop run in the presence of empty results.dat files 
                       (which can happen if there are problems with disk access
                       in a cluster run).
                   JA: Allow reading up to 5M weighted events in combine_events.

1.4.8.2 (30/07/12) OM: Allow AE(1,1), AE(2,2) to not be present in the SLHA1 card
                       (1.4.8 crashes if they were not defined in the param_card)
                   OM: Add a button Stop-job for the cluster and make nicer output
                       when the user presses Ctrl-C during the job.

1.4.8 (24/07/12)  JA: Cancel running of integration channels where the BW
                      structure makes it impossible to get any events. This
                      can significantly speed up event generation for processes
                      with conflicting BWs.
                  JA: Minor modification of integration grid presetting in
                      myamp.f, due to the above point.
                  JA: Raise exception if a decay process has decaying particles
                      that are not present in the corresponding core process
                      (this might help avoid syntax mistakes).
                  JA: Fixed subprocess group combination also for the case
                      when different process flags @N are given to different
                      decays of the same core process (sorry, this was missed
                      in v. 1.4.7).
                  JA: Fixed crash for process p p > w+ w+ j j t t~ when all 
                      w and t/t~ are decayed (bug #1017912, thanks to Nicolas
                      Deutschmann).
                  JA: Fixed array dimension for diagrams with a single s-channel
                      propagator (caused crash for certain compilers, bug #1022415
                      thanks Sho Iwamoto).
                  JA: Fixed crash for identical decay chains for particle-anti- 
                      particle when only one of the two is decayed, introduced 
                      in v. 1.4.7 (thanks John Lee).
                  OM: Ensure that matching plots are replaced correctly when
                      Pythia is run reusing a tag name.
                  OM: Improved check for YE/AE, YU/AU, YD/AD for first two
                      generations in SLHA1<->2 converter (thanks Abhishek).

1.4.7 (25/06/12)  JA: Change the random seed treatment to ensure that the original 
                      seed is stored in the banner (this was broken in v. 1.4.0).
                      If a non-zero seed is given in the run_card, this seed
                      is used as starting value for the SubProcesses/randinit file,
                      while the seed in the run_card is set to 0.
                      This way, the seed for a multi_run is automatically
                      updated in the same way as for individual runs.
                  TS + JA: Fix problem with duplicate random seeds above 64000.
                      Now, random seeds up to 30081*30081 can safely be used.
                  JA: Turn off automatic checking for minimal coupling orders
                      in decay processes A > B C ...
                  JA: Ensure that automatic coupling order determination works
                      also for effective theories with multiple orders in an
                      interaction (thanks Claude and Gizem Ozturk).
                  JA: Optimize phase space integration and event generation
                      for decay processes with very squeezed mass spectrum.
                  JA: Ensure that identical matrix elements in different process 
                      definitions are combined also when using the decay chain 
                      formalism (thanks to Zhen Liu for pointing this out).
                  BF+JA: Updated the NMSSM model to the latest FR version.
                  OM: Change EW_dim6 to remove all interactions which don't 
                      impact three boson scattering.
                  JA: Fixed problem in matrix element combination which allowed
                      non-identical matrix elements to be combined in certain
                      complicated processes (such as p p > l vl l vl l vl),
                      resulting in lines with Z > e+ mu- in the event file
                      (bug #1015032, thanks Johannes E for reporting).
                  JA: Fixed minor typo in myamp.f.
                  OM: Fixed minor behavior restriction of multi_run (thanks to
                      Joachim Kopp).
                  OM: Improved condor cluster support when the cluster is 
                      unresponsive (should avoid some crashes on the web).
                  JA: Fixed support for color sextets in addmothers.f
                      (thanks Nicolas Deutschmann for reporting).          
                  JA: Make sure that also the SubProcesses directories are 
                      cleaned when running bin/compile in a gridpack.
                  JA: Removed the confusing makefile in Template and replace it
                      with scripts to create madevent.tar.gz and gridpack.tar.gz.
                  
1.4.6 (16/05/12)  JA: Added cuts on lepton pt for each of the 4 hardest leptons
                  OM: Allow bin/madevent script to be run with a single line command
                      example ./bin/madevent multi_run 10 
                  OM: Adding the 4 higgs interactions in the standard model UFO model
                  JA: Added new restriction card for the sm model with massive
                      muon and electron, and non-zero tau decay width
                  JA: Ensure assignment of colors to intermediate propagators
                      works also in fermion flow- and color flow-violating 
                      RPV processes (thanks Brock Tweedie for finding this).
                  JA: Fix crash for certain fermion flow violating decay chains
                      (introduced in v. 1.3.27) (again thanks to Brock Tweedie).
                  JA: Fix crash for decay chains with multiple decays involving 
                      the same particles (thanks Steve Blanchet for reporting)
                  JA+OM: Fix crash for Pythia8 output with multiparticle vertices
                      (thanks to Moritz Huck for reporting this.)
                  OM: Fixing ALOHA output for C++/python.
                  OM: Fix a crash occurring when trying to create an output on 
                      an existing directory (thanks Celine)

1.4.5 (11/04/12)  OM: Change the seed automatically in multi_run. (Even if the seed
                      was set to a non automatic value in the card.)
                  OM: correct a minor bug #975647 (SLHA convention problem) 
                      Thanks to Sho Iwamoto
                  OM: Improve cluster support (more secure and complete version)
                  JA: Increased the number of events tested for non-zero helicity
                      configurations (needed for goldstino processes).
                  OM: Add a command to remove the file RunWeb which was not always
                      deleted correctly
                  OM+JA: Correct the display of number of events and error for Pythia 
                     in the html files.
                  OM: Changed the way the stdout/stderr are treated on the cluster
                      since some clusters can't support having the same output file
                      for both. (thanks abhishek)

1.4.4 (29/03/12)  OM: Added a command: "output aloha" which allows to create a
                      subset (or all) of the aloha routines linked to the
                      current model
                  OM: allow to choose the duration of the timer for the questions.
                      (via ./input/mg5_configuration.txt)
                  OM: Allow UFO model where G is not defined.
                  OM: allow to use ~,~user, ${var} in the path. Improve support
                      for path containing spaces.
                  JA: Fixed LHAPDF functionality which was broken in v. 1.4.0
                  JA: Allow non-equal mixing angles in mssm restrict cards
                      (as needed for cards from some spectrum generators)
                  JA: Fixed script addmasses.py for complicated events such as
                      p p > t t~ + jets with decays of t and t~.
                  JA: Added GE cluster to the list in cluster.py.
                  JA: Allow up to 1M events in a single run. Note that the 
                      unweighting (combine events) step gets quite slow with
                      so many events. Also note that if Pythia is run, still
                      maximum 50k events is recommended in a single run. 
                  OM: Fix a problem linked to the filesystem which makes new files
                      non-executable by default. (bug #958616)
                  JA: Fixed buffer overflow in gen_ximprove when number of
                      configs > number of diagrams due to competing resonances
                      (introduced in v. 1.4.3).

1.4.3 (08/03/12)  JA: Reintroduced the possibility to completely forbid
                      s-channel diagrams, using the $$ notation. Note that
                      this should be used with great care, since the result
                      is in general not gauge-invariant. It is in general
                      better to use the $ notation, forbidding only onshell
                      s-channel particles (the inverse of decay chains).
                  JA: Automatically ensure that ptj and mmjj are below xqcut
                      when xqcut > 0, since ptj or mmjj > xqcut ruins matching.
                  OM: Add LSF to the list of supported clusters (thanks to Alexis).
                  OM: change the param_card reader for the restrict file.
                      This allows to restrict models with 3 lha ids (or more)
                      (thanks to Eduardo Ponton).
                  OM: forbids to run 'generate events' with python 2.4.
                  OM: Include the configuration file in the .tar.gz created on 
                      the web (thanks to Simon) .
                  OM: Fix a Mac specific problem for edition of Delphes card.
                      (thanks to Sho Iwamoto).
                  OM: ALOHA modifications:
                       - Change sign convention for Epsilon (matching FR choices)
                       - For fermion vertices, forces that _1 always returns the
                         incoming fermion and _2 returns the outgoing fermion.
                         (This modifies conjugate routine output)
                       - Change the order of argument for conjugate routine
                         to expect IO order of fermion in all cases.
                       Note that the two last modifications match MG5 conventions
                       and that those modifications correct bugs for interactions
                       a) subject to conjugate routine (i.e. if the model has 
                          majorana)                       
                       b) containing fermion momentum dependencies in the Lorentz
                          structure  
                       All model included by default in MG5 (in particular sm/mssm)
                       were not affected by those mismatches of conventions.
                       (Thanks to Benjamin Fuks)
                  OM: make acceptance test more silent.  
                  OM: return the correct error message when a compilation error occurs.
                  OM: some code re-factoring.

1.4.2 (16/02/12) JA: Ensure that matching works properly with > 9 final state
                      particles (by increasing a buffer size in event output)
                 OM: add a command "import banner" in order to run a full run
                      from a given banner.
                 OM: Fix Bug #921487, fixing a problem with home-made models
                      in the definition of Particle/Anti-Particle. (Thanks Ben)
                 OM: Fix a formatting problem in me5_configuration.txt 
                      (Bug #930101) Thanks to Arian
                 OM: allow to run ./bin/mg5 BANNER_PATH and
                      ./bin/mg5 PROC_CARD_V4_PATH
                 OM: Various small fixes concerning the stability of the html 
                      output.
                 OM: Change the server used to download td, since cp3wks05 has a
                      hard disk failure.

1.4.1 (06/02/12) OM: Fix the fermion flow check which was wrongly failing on
                      some models (Thanks to Benjamin)
                 OM: Improve run organization efficiency (which speeds up the 
                      code on cluster) (Thanks to Johan)
                 OM: More secure html output (Thanks to Simon)

1.4.0 (04/02/12) OM: New user interface for the madevent run. Type:
                      1) (from madevent output) ./bin/madevent
                      2) (from MG5 command line) launch [MADEVENT_PATH] -i
                      This interface replaces various scripts like refine,
                      survey, combine, run_..., rm_run, ...
                      The script generate_events still exists but now calls
                       ./bin/madevent. 
                 OM: For MSSM model, convert param_card to SLHA1. This card is
                      converted to SLHA2 during the MadEvent run since the UFO 
                      model uses SLHA2. This allows to use Pythia 6,
                      as well as having a coherent definition for the flavor.
                 JA+OM: For decay width computations, the launch command in 
                      addition to compute the width, creates a new param_card 
                      with the width set to the associated values, and with the 
                      Branching ratio associated (useful for pythia). 
                 NOTE: This param_card makes sense for future runs ONLY if all
                      relevant decays are generated.
                 EXAMPLE: (after launch bin/mg5):
                       import model sm-full
                       generate t > b w+
                       define all = p b b~ l+ l- ta+ ta- vl vl~
                       add process w+ > all all
                       add process z > all all
                       define v = z w+ w-
                       add process h > all all
                       add process h > v v, v > all all
                       output
                       launch
                 OM: change output pythia8 syntax: If a path is specified this 
                      is considered as the output directory.
                 OM: Change the path of the madevent output files. This allows 
                      to run pythia/pgs/delphes multiple times for the same set
                      of events (with different pythia/... parameters).
                 OM: Madevent output is now insensitive to the relative path
                      to pythia-pgs, delphes, ... In consequence you don't need
                      anymore to have your directory at the same level as 
                      Template directory. 
                 OM: MadEvent checks that the param_card is coherent with the 
                      restriction used during the model generation. 
                 OM: Model restrictions will now also force opposite numbers to
                      match (helpful for constraining to a rotation matrix).
                 OM: Change the import command. It's now allowed to omit the
                      type of import. The type is guessed automatically.
                      This is NOT allowed on the web.
                 OM: Add a check that the fermion flow is coherent with the 
                      Lorentz structure associated to the vertex.
                 OM: Add a check that the color representation is coherent. 
                      This allows to detect/fix various problems linked
                      to some new models created by FR and SARAH.
                 OM: Change the default fortran compiler to gfortran.
                 OM: Add the possibility to force which fortran compiler will
                      be used, either via the configuration file or via the set 
                      command.
                 OM: Add the possibility to bypass the automatic opening of 
                      the web browser (via the configuration file: 
                      ./input/mg5_configuration.txt )
                 OM: add 'save options' command to save the current configuration 
                      in the configuration file. 
                 OM: Change the scheme of questions when running madevent and 
                      allow to specify in the command interface if you
                      want to run pythia/pgs/...
                      Allow to put the answers to the questions in the 
                      proc_card.dat.
                 OM: Add options for the display command:
                      a) display options: return the current option value. 
                        i.e. those set via the set command and/or via the 
                        configuration file
                      b) display variable NAME: return the current string 
                        representation of NAME and/or self.NAME .
                      c) display coupling_order: return the coupling orders with
                        their associated weight (for automatic order restriction)
                      d) display couplings now returns the list of all couplings
                        with the associated expression
                      e) display interactions [PART1] [PART2] [PART3] ...
                         display all interactions containing the particles set
                         in arguments 
                 OM: New Python script for the creation of the various html pages.
                      This requires less disk access for the generation of the files.
                 OM: Modify error treatment, especially for Invalid commands
                      and Configuration problems.
                 JA: Ensure that we get zero cross section if we have
                      non-parton initial states with proton/antiproton beams
                 OM: Improve cluster support. MadEvent now supports PBS/Condor/SGE
                      Thanks to Arian Abrahantes for the SGE implementation.
                 OM: Improve auto-completion (better output/dealing with multi line/...)
                 OM: Improve the parallel suite and change the release script to run
                      some of the parallel tests. This ensures even higher stability 
                      of the  code for the future releases.
                 JA: Changed the way gridpacks work: Set granularity to 1
                      (so randomly select channels only if they should generate 
                      less than 1 event), but allowing channels to run down to a single
                      iteration. This removes all old problems with increased
                      variance for small channels in the gridpacks, while giving 
                      even faster event generation.

                 Thanks to Johan Alwall, Sho Iwamoto for all the important 
                 testing/bug reports.


1.3.33 (01/01/12) JA: Revisited colors for propagators in addmothers.f
                      to ensure that propagators in color flow
                      violating processes get the correct color
                      from initial state particles (thanks to
                      Michele Gabusi for forcing me to do this).

1.3.32 (21/12/11) JA: Fixed a bug in the PDF reweighting routine,
                      which caused skewed eta distributions for
                      matched samples with pdfwgt=T. Thanks to Giulio
                      Lenzi for finding this.
 
1.3.31 (29/11/11) OM: Fix an overflow bug in RAMBO (affects standalone
                     output only)
                  PdA (via OM): Change RS model (add a width to the spin2)
                  OM: Fix a bug in the cuts associated to the allowed mass of all
                      neutrinos+leptons (thanks to Brock Tweedie for finding it)
                  OM: Remove some limitation in the name for the particles


1.3.30 (18/11/11) OM: Fix a bug for the installation of pythia-pgs on a 64 bit
                      UNIX machine.
                  OM: If ROOTSYS is defined but root is not in the PATH, add it
                      automatically in create_matching_plots.sh
                     This is required for the UIUC cluster.

1.3.29 (16/11/11) OM: Fixed particle identities in the Feynman diagram drawing
                  JA: Fixed bug in pdf reweighting when external LHAPDF is used.
                  OM+JA: Simplify the compilation of pythia-pgs package.


1.3.28 (14/11/11) OM+JA: Fix special case when Lorentz structure combining
                      two different Majorana particles depends on the
                      incoming/outgoing status of the Majorana particles
                      (needed for MSSM with Goldstino).
                  JA: Fixed problem with colors in addmothers.f for complicated
                      multiparticle vertices and simplified color treatment 
                      (thanks to Gauthier Durieux for pointing this out).
                  JA: Further improved gridpack parameters
                  OM: Update the parallel test (now testing against MG5 1.3.3)
                  OM: Include some parallel test in the release script.


1.3.27 (05/11/11) JA: Fix bug in mirrored amplitudes (sometimes
                      amplitudes that should not be flagged as
                      mirrored were flagged as mirrored). Thanks
                      Marco Zaro for reporting this!
                  JA: Fix another problem getting enough events in
                      gridpack mode (it was not completely fixed in
                      v. 1.3.24). Thanks Alexis!
                  JA: Added "!" comments for all parameters in the default
                      run_card, since apparently this is still needed
                      for g77 to correctly read the parameters.
 
1.3.26 (31/10/11) JA: Fix color setting in MadEvent event file for
                      multiparticle vertices, which was not taken into
                      account in the upgrade in v. 1.3.18
                  OM: Fixed mmnl cut (inv. mass of all leptons and neutrinos)
                      which was never active.
                  OM: Fix the td install on Linux where a chmod was missing

1.3.25 (27/10/11) JA: Ensure that the correct intermediate resonance
                      is always written in the event file, even when we
                      have resonances with identical properties.
                  OM: Fix the bug forcing to quit the web browser in order to
                      have MG5 continuing to run.
                  OM: Change the tutorial in order to allow open index.html
                      after the output command. 

1.3.24 (22/10/11) JA: Fix problem with getting enough events in gridpack
                      mode (this was broken in v. 1.3.11 when we moved
                      from events to luminosity in refine). Thanks to
                      Alexis Kalogeropoulos.

1.3.23 (19/10/11) JA: Allow user to set scales using setscales.f again 
                      (this was broken in v. 1.3.18). Thanks to Arindam Das.
                  JA: Ensure that the error message is displayed if the
                     "make" command is not installed on the system.
 
1.3.22 (12/10/11) JA: Fixed another bug (also introduced in 1.3.18), which 
                      could give the wrong ordering between the s-channel 
                      propagators for certain multiprocess cases (this
                      also led to a hard stop, so don't worry, if you get
                      your events, the bug doesn't affect you). Sorry about
                      that, this is what happens when you add a lot of
                      new functionality...

1.3.21 (12/10/11) OM: Add a new command: install.
                      This allows to install quite easily different packages
                      developed for MadGraph/MadEvent. The list of available
                      packages is pythia-pgs/MadAnalysis/ExRootAnalysis/Delphes
                  OM: Adding the TopEffth model
                  OM: Improve display particles and autocompletion in
                      presence of nonpropagating particles
                  OM: Fix Aloha bug linked to four fermion operator
                  PA: fix the problem of degenerate color basis in the
                      diquark sextet model
                  JA: Fixed bug in cluster.f that created a hard stop,
                      introduced in 1.3.18.

1.3.20 (09/10/11) JA: Fixed bug in myamp.f that created a hard stop
                      error for certain cases with many processes with
                      different propagators in the same subprocess dir.

1.3.19 (06/10/11) JA: Fixed problem with SubProcesses makefile on Linux,
                      introduced in 1.3.18.

1.3.18 (04/10/11) JA: Use model information to determine color of particles
                      for reweighting and propagator color info.
                  JA: Changed the definition of "forbidden s-channels"
                      denoted by "$" to exclude on-shell s-channels while
                      keeping all diagrams (i.e., complementary to the decay
                      chain formalism). This reduces the problems with 
                      gauge invariance compared to previously.
                      "Onshell" is as usual defined by the "bwcutoff" flag 
                      in the run_card.dat.
                  JA: Enable proper 4-flavor matching (such as gg>hbb~+jets)
                      Note that you need the Pythia/PGS package v. 2.1.9 or 
                      later to use with 4F matching.
                      Changes include: alpha_s reweighting also for b vertices,
                      new scale treatment (mu_F for pp>hbb~ is (pT_b^max*m_Th)),
                      no clustering of gluons to final-state massive particles
                      in MadEvent.
                  JA: Ensure that factorization scale settings and matching works
                      also in singlet t-channel exchange processes like
                      single top and VBF. The dynamic factorization
                      scale is given by the pT of the scattered quark
                      (on each side of the event).
                Note: You need the Pythia/PGS package v. 2.1.10 or later
                      to use with VBF matching, to ensure that both radiated
                      and scattered partons are treated correctly
                      - scattered partons need to be excluded from the matching,
                      since their pT can be below QCUT. An even better
                      treatment would require to individually shower and match
                      the two sides in Pythia, which is not presently possible.
                Note: In the matched 4-flavor process p p > t b~ j $ w+ w- t~ +
                      p p > t b~ j j $ w+ w- t~, there is an admixture
                      of t-channel single top (with up to 1 radiated jet) 
                      and s-channel single top (with up to 2 radiated jets). 
                      In this case, the automatic determination of maximum 
                      multiplicity sample doesn't work (since max in the file 
                      will be 2 jets, but for t-channel max is 1 jet).
                      So MAXJETS=1 must be specified in the pythia_card.dat.
                  JA: Fixed pdf reweighting for matching, which due to a mistake
                      had never been activated.
                  JA: Improved phase space integration presetting further by 
                      taking into account special cuts like xpt, ht etc.
                  JA: Introduce new convention for invariant mass cuts
                      - if max < min, exclude intermediate range
                      (allows to exclude W/Z dijet resonances in VBF processes)

1.3.17 (30/09/11) OM: Fix a crash created by ALOHA when it tries to create the full
                      set of ALOHA routines (pythia8 output only).

1.3.16 (11/09/11) JA: Fixed the problem from 1.3.12.

1.3.15 (09/09/11) OM: remove the fix of 1.3.12
                      (No events in output for some given processes)

1.3.14 (08/09/11) OM: Fix a bug in the RS model introduced in 1.3.8

1.3.13 (05/09/11) JA: Fixed bug with cut_decays=F which removed cuts also for
                      non-decay products in certain channels if there is
                      a forced decay present. Note that this does not affect
                      xqcut, only pt, minv and eta cuts.
                  JA: If non-zero phase space cutoff, don't use minimum of
                      1 GeV (this allows to go to e.g. 2m_e invariant mass for
                      \gamma* > e+ e-).

1.3.12 (01/09/11) JA: Fixed problem with decay chains when different decays
                      result in identical final states, such as
                      p p > go go, (go > b1/b1~ b/b~, b1/b1~ > b/b~ n1)
                      (only one of the decay chains was chosen, instead of
                      all 3 combinations (b1,b1), (b1,b1~), (b1~,b1~))
                  JA: Allow for overall orders also with grouped subprocesses
                  JA: Ensure that only leading color flows are included in event
                      output (so no singlet flows from color octets).
                  JA: Fixed small bug in fermion flow determination for multifermion
                      vertices.

1.3.11 (26/08/11) JA: Improved precision of "survey" by allowing 4th and 5th 
                      iteration if accuracy after 3 iterations < 10%.
                  JA: Subdivide BW in phase space integration for conflicting BWs 
                      also for forced decays, to improve generation with large
                      bwcutoff in e.g. W+ W- production with decays.
                  JA: Do refine using luminosity instead of number of events,
                      to work with badly determined channels.
                  JA: Don't use BW for shat if mass > sqrt(s).
                  JA: Fixed insertion of colors for octet resonances decaying to 
                      octet+singlet (thanks Bogdan for finding this)

1.3.10 (23/08/11) OM: Update ALOHA version
                  OM: increase the waiting time for jobs to physically write the results on
                      the disks (in order to reduce trouble on the cluster).

1.3.9 (01/08/11)  OM: Add a new model DY_SM (arXiv:1107.5830). Thanks to Neil 
                      for the generation of the model 

1.3.8 (25/07/11)  JA: Replace the SM and HEFT models with latest versions using
                      the Wolfenstein parameterization for the CKM matrix.
                  JA: Implemented reading of the new UFO information about
                      coupling orders (order hierarchy and expansion_order).
                  JA: New "coupling order" specification WEIGHTED which checks
                      for  sum of coupling orders weighted by their hierarchy.
                  JA: Implemented optimal coupling orders for processes from any
                      model if no coupling orders specified.

1.3.7 (21/07/11)  JA: Fix makefiles for some v4 models that were forgotten
                      in v. 1.3.5

1.3.6 (18/07/11)  OM: Ensure that the new makefiles work on the web

1.3.5 (14/07/11): JA: New organization of make files, ensure that compilation works 
                      for all modes (with/without LHAPDF, static/dynamic, 
                      regular/gridpack) for both Linux and Mac OS X (be careful with 
                      dynamic libraries on Mac OS X though, since it seems that 
                      common blocks might not work properly)
                  JA: Fixed proper error messages and clean stop for compilation 
                      errors during MadEvent run.

1.3.4 (05/07/11): OM: More informative error message when a compilation error occurs

1.3.3 (29/06/11): JA: Fixed diagram symmetry for case when there are
                      no 3-vertex-only diagrams
                  JA (by OM): More informative error when trying to generate invalid 
                      pythia8 process

1.3.2 (14/06/11): OM: Fix fortran output when a model is case sensitive
                        (Bug if a coupling was depending on a case sensitive parameter)
                  SdV: Remove an annoying print in the new cuts (added in 1.3.0)
                  OM: Fix a compilation problem in the standalone cpp output

1.3.1 (02/06/11): JA: Fixed missing file bug with the introduction of
                      inclusive HT cut

1.3.0 (02/06/11): JA: Allow for grouped subprocesses also for MG4 models
                  JA: Improved multiprocess diagram generation to reuse
                      diagrams for crossed processes
                  JA: Automatic optimization of order of particles in
                      multiparticle labels for optimal multiprocess generation
                  JA: Improved efficiency of identification of identical
                      matrix elements
                  JA: Improved identification of diagrams with identical
                      divergency structure for grouped subprocesses
                  JA: Included more fine-grained run options in the
                      run_card, including helicity summation options,
                      whether or not to set ptj and mjj automatically
                      based on xqcut, etc.
                  JA: Fixed some minor array limit and arithmetics warnings
                      for extreme decay and decay chain processes.
                  SdV: Added cuts on H_T(all jets, light and b)
                  OM: Fixed minor bug related to cluster option in launch

1.2.4 (15/05/11): JA: Fixed long-standing bug in DECAY relating to
                      the cross section info in <init> block, and
                      fixed parameter reading for MG5 SM model.

1.2.3 (11/05/11): JA: Fixed problem with scale choice in processes with mixed 
                      QED/QCD orders, e.g. p p > t t~ QED=2. Note that this fix
                      doesn't work for p p > t t~ j j QED=4 which should still
                      be avoided.
                  JA: Added the ptllmin/max options in the default run_card.dat

1.2.2 (09/05/11): OM: fix ALOHA symmetries creating not gauge invariant result 
                      for scalar octet

1.2.1 (08/05/11): OM: reduce the quantity of RAM used by matrix.f
                  OM: support the speed-up from psyco if this python module is installed
                  OM: fix a minor bug in the model parsing
                  OM: add the check of valid model.pkl also for v4 model
                  OM: add a check that UpdateNotes is up-to-date when
                      making a release
                  JA: Fixed problem in phase space generation for
                      s-channel mass > s_tot

1.2.0 (05/05/11): OM: minor fixes on check charge conjugation
                  OM: add a check on the path for the validity of the model.pkl
                  JA: Fixed problem with combine_runs on certain compilers

1.1.2 (03/05/11): OM+JA: Fixed problem for models with multiple
                      interactions for the same set of particles,
                      introduced in v. 1.1.1
 
1.1.1 (02/05/11): JA: Replaced (slow) diagram symmetry determination by
                      evaluation with fast identification based on diagram tags.
                  JA: Replacing the "p=-p" id=0 vertex produced by diagram 
                      generation algorithm already in the diagram generation,
                      simplifying drawing, helas objects and color.
                  JA: Fixed compiler warnings for unary operator.
                  JA: Always set all coupling orders for diagrams
                      (needed for NLO implementations).
                  OM: Improved and more elegant "open" implementation for
                      the user interface.
                  OM: minor fixes related to checking the gauge

1.1.0 (21/04/11): JA: Removed hard limit on number of external particles in 
                      MadEvent, allowing for unlimited length decay chains there
                      (up to 14 final state particles successfully integrated).
                  JA: Improved helicity selection and automatic full helicity 
                      sum if needed. Optimization of run parameters.
                  JA: New flag in run_card.dat to decide whether basic cuts
                      are applied to decay products or not.
                  OM: Merged ALOHA calls for different lorentz structures 
                      with the same color structures, increasing the speed and 
                      efficiency of matrix element evaluations.
                  OM: Added new "open" command in command line interface,
                      allowing to open standard file types directly.
                      Automatically open crossx.html at launch.
                  JA: Fixed MadEvent bugs for multiparton processes with 
                      conflicting decays and some faulty array limits.
                  JA: Suppressed scary but irrelevant warnings for compiling 
                      2->1 and 1->2 processes in MadEvent.
                  JA: Pythia 8 output further optimized.
                  JA, OM: Several minor fixes relating to user interface etc.

1.0.0 (12/04/11): Official release of MadGraph 5. Some of the features:
                  - Complete FeynRules compatibility through the UFO interface
                  - Automatic writing of HELAS routines for any model in
                    Fortran, C++ or Python through ALOHA
                  - Matrix element output in Fortran, C++ or Python
                  - Output formats: MadEvent, Pythia 8, Standalone (Fortran/C++)
                  - Support for spin 0, 1/2, 1, 2 particles
                  - Support for color 1, 3, 6, 8
                  - Revamped MadEvent with improved subprocess directory 
                    organization and vastly increased speed and stability
                  - Unlimited length decay chains (up to 12 final state
                    particles tested with MadEvent, see v. 1.0.1)
                  - Process checks for new model implementations
                  - ...and much more (see paper "MadGraph 5: Going Beyond")