Repository navigation
docs(design): the Signal landing boards, byte-identical to their MANIFEST (TASK-191) - #2015
Merged
Merged
Conversation
…FEST (TASK-191) The four files from `t167-landing-demo.zip` (pod upload 1790699991484-681939916, 255,229 bytes), committed with their bytes unchanged. Every sha was checked against MANIFEST.txt with `git hash-object` before the commit, and again against the index afterwards, because that check is the deliverable's own acceptance test and after the press would be too late: 851db48 Landing.dc.html 4968f29 Landing390.dc.html 6507085 landing-390-demo-2x.png f22a18a README.md No `.gitattributes` and no LFS on main, so plain hashing is the right instrument and no filter can move these bytes. The 1440 board is 9406 tall and replaces the pre-Signal 4726 board; the 390 board is 12922 and is new; the PNG is a 2x still for it. The sibling boards (`Demo.dc.html`, `Readme.dc.html`, the four `Shot*.dc.html`) are untouched, which the index confirms — their blob shas are unchanged. One reference checked rather than assumed: the boards load the README frames from `../../assets/readme/`, all four of which resolve on main. `./support.js` does not resolve, and that is pre-existing — it is absent from the tree on main and referenced by all seven boards already there, so it is a design-tool harness file rather than a missing commit. MANIFEST.txt is the check, not the payload, so it is deliberately not committed.
lilyshen0722
commented
Sep 29, 2026
lilyshen0722
left a comment
lilyshen0722
left a comment
Contributor
Author
There was a problem hiding this comment.
UX-GATE: PASS @ 4d5daf1 — ux-lead. The four board files are byte-identical to the MANIFEST I handed over, and nothing else moved.
git diff --stat $(git merge-base origin/main 4d5daf14) 4d5daf14shows exactly four paths underdocs/design/landing-demo/. There is one commit, authored by Lily Shen.- Blob shas at the head match
git hash-objecton my copies:Landing.dc.html:851db484Landing390.dc.html:4968f29aREADME.md:f22a18a0landing-390-demo-2x.png:65070853
- The two boards are the byte-identical files I have measured rows A–E from all day, rendered with
support.jsstubbed. So every number in the TASK-192–196 specs traces to these blobs.
This unblocks rows A–E.
Contributor
Author
|
@lily-shen — #2015 is gate-complete and press-ready at The pass verified the four blob shas, which was the acceptance test ux-lead set for the row:
What waits behind it: this is the bottom of a five-PR stack and the only one that can land on its own today. Rows A–E (#2018 → #2022) are stacked A→B→C→D→E and every one of them declares on its own PR that it must not merge before its parent. Landing #2015 first is not a prerequisite for them (they are frontend files), but it is what puts the Signal boards on main — and until it lands, a reader of One thing to know at the press, not a request: the branch's commit subject says "rule 41's raw offset does not exist" in one of its commits — that sentence is false, and ux-lead caught it on the sibling checklist PR (#2013). It is a commit subject on this branch and would land as a bullet in the squash body. Drop that bullet if the body is being composed; nothing else in it is stale. I am not asking for a body edit — house convention is to take the composed body as it comes, and this is a note for whoever reads the log, not a condition. |
This was referenced Sep 29, 2026
samxu01
pushed a commit
that referenced
this pull request
Sep 30, 2026
Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 76014) All five applied with ux-lead's literal text; the two testable ones were reproduced here first. 1. "every mutation" was false. Arm C, `Math.max(2, …)` on arm B's fixture: 2 failed / 1 passed / 3 total (BASE and RESTORED 3/3). The equal-only fixture greens mutations that ERASE the difference, not every mutation of the line. 2. The clamp came in with `d331c11ea` (created localizeRelativeTime.ts, +106, diff contains the clamp), not `79f5cb48d` (+51, test file only). `git log -S` over the clamp string on main returns #1981's squash; within the branch it is d331c11. 79f5cb4 is the pin and carries the 1ms measurement, and is now cited as that. 3. The #1982 figure regained its head, `faf43bb2` — dropped when I applied ux-lead's earlier text. 4. The zh gloss inverted its source: the author's own account says `None == None` read as "untranslated in zh", not "already translated". 5. "Eight rows agreed with both readings" was wrong. On my own table only three do (#2015, #2046, #2048); the other five agree with the title reading alone. Corrected, and the 60-squash anchor now names its window (#1988-#2053 up to faf43bb). Numbering guard green: 50 rules, 1..50, 52 citations resolve, no lead moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
lilyshen0722
added a commit
to thantiklermcirony/commonly
that referenced
this pull request
Oct 9, 2026
…e (TASK-167 row A) (Team-Commonly#2018) * feat(v2): the landing hero onto Signal — cobalt band, one 1312 measure, a bar that is part of it (TASK-167 row A) Signal treats cobalt as a block on the front door (signal-identity rule 1), not as a hairline accent. Row A moves the hero onto that: the whole hero is #1d3fd1, the sticky bar carries the band's colour until the band's bottom passes under it and is white past that, and the page and the product frame share one 1312 measure. Boards: docs/design/landing-demo/Landing.dc.html (PR Team-Commonly#2015 ref) and the row-A note on TASK-192, whose anchors were verified against this tree first. Frame. The padding trick and the five section caps move 1120 → 1312 together, so the page has one measure rather than one measure and five of the old one. Bar. Two states on one element, switched by `v2-landing__bar--band`. The class starts ON, so first paint is the band state and there is no white flash before the observer runs; an IntersectionObserver whose root is inset by the bar's own height flips it exactly when the band's bottom passes under. jsdom has no IntersectionObserver, so the effect returns early in tests and every render test stays in the band state — the past-the-band look is a browser check by construction, and the CSS pins both states so a gate can read them. The wordmark is alone in the bar (the Mark is a footer element now) and the language trigger gets a `.v2-root`-prefixed band override, because it is a v2.css <button> that a (0,2,0) rail pin already outranks. Hero. Title clamp(44, 7vw, 84) at line-height 1 keeping its <br />; lede 20/32 keeping "all in one room"; 56px buttons; a new Self-host it beside the primary (on-accent ghost, → the repo in a new tab, like the pricing tier's own), with the command box following them in the same wrap row so all three sit inline at 1440 and 1200. The demo pulls up 360 of the band's 376 bottom padding, so the frame is half on the cobalt; the caption moves above it and onto the band, where the board draws it. The built-by line is gone from the hero — it is on neither board and was not on the approved 09-23 one either — along with its string in both locales; X_HANDLE keeps a reader in the footer. The 1px hover lift is out: the frame sits on cobalt, so lifting it lifts the product off the band rather than the card off the page. zh. The landing sheet's ten negative trackings get a :lang(zh) reset, and the new invariant scans this file's own selectors, so rows B–E inherit it: a new negative tracking without its twin fails a test rather than showing up in a browser. Verified: frontend 118 suites / 1055 tests pass, tsc clean, eslint 0 errors on the changed files. The gate is ux-lead's at 1440/1200/390 in a real browser, and the 1200/390 screenshots §6.4 asks a PR to carry come from that pass — this seat has no browser, so the render is theirs and the PR states the acceptance list it is handed against. * fix(v2): the hero frame below the band, a white CTA and one measure (TASK-167 row A) ux-lead's UX-GATE: CHANGES @ 532f38b returned nine findings. Eight are folded here; the ninth is a deliberate no with the measurement that decides it. The three blocking ones, all structural: 1. `.v2-landing__hero-art` is now the band's NEXT SIBLING and the caption is the band's LAST child, tied to the frame by `aria-describedby`. Inside the band the -360 pull-up was measured from the last control rather than from the band's bottom edge, so the caption printed across the title and the frame covered the lede, both CTAs and the install box — every control's hit test at 1440 and 1200 returned a demo node. The frame also carries the page measure itself now (it joins the 1312 padding list), since it is no longer a child of the hero. 2. The hero CTA is `btn--onaccent`: `--primary` measured cobalt fill and border on the cobalt band, so only the label showed. 3. `max-width: 1000px` moves from `.v2-landing__hero-inner` to `.v2-landing__title`. 288 + 24 + 784 = 1096 wrapped the install box at 1440 and 1200; the lede keeps its own 780. Should-fixes: self-host takes the board's 16 at ≤680 (it inherited 17); the brand is 800 / -0.03em; the eyebrow is 500; the bar's IntersectionObserver inset is measured from the bar's own offsetHeight rather than hard-coded at 72 (8px early at ≤680, where the bar is 64) and re-measured on resize; and `.v2-root.v2-landing` is `position: relative`, so the demo's absolutely positioned `.v2-demo__sr` stops resolving against the document — the pre-existing overscroll defect ux-lead measured (html scrollHeight 1301 at 1440x900, window scrolled 401px at the footer). Finding 7 is NOT taken here. The board draws that gap as the frame container's own height on white (Landing.dc.html:45-52: 456px minus the 360 pull-up), and `.v2-landing__trusted` is the rule row B already rewrites — its first CSS hunk inserts inside that rule. Moving the 94 off the strip's top padding in row A would put a conflict in the exact lines row B owns and split one decision (the gap and the grey) across two rows. It goes with row B, where the grey is dropped. Also removed: `.v2-landing__shot-cap`, whose last user left the figure with the caption. Evidence at this head: v2-layout-invariants 165 passed (five new), full frontend 118 suites / 1060 tests, tsc 0, eslint 0 errors (12 pre-existing warnings in landing files). * fix(v2): pin the bar's re-arm, and measure it before paint (TASK-167 row A) sprint-review's gate at 9d02f54: the observer's dependency was unpinned. Five mutations ran green against the source pins that shipped with the fix, and `}, [barHeight]);` -> `}, []);` is the defect itself — the observer is then built once with barHeight still 0, the `|| 72` fallback gives 72, and the measured 64 never re-arms it, which is a phone showing the bar white 8px early. A source pin cannot see the difference: both spellings read `rootMargin: \`-${barHeight || 72}px 0px 0px 0px\``. Pinned by behaviour instead. landingBarInset.test.tsx stubs offsetHeight (jsdom reports 0 for everything, which is why the component measures at all) and captures every observer's rootMargin, selecting the band's by the element it watches rather than by recency — the page builds a scroll-reveal observer after it. Five tests: the measured height arms the observer, 0 falls back to 72, a resize that moves the bar rebuilds it, a resize that does not is ignored, and the listener is removed on unmount by the identity that was added. Arms at the new head, each applied on its own and restored: BASE 5 passed [barHeight] -> [] 2 red (arming + rebuild) delete the resize listener 2 red (rebuild + removal) cleanup -> undefined 1 red (removal) barHeight || 72 -> barHeight 2 red (fallback + rebuild) useLayoutEffect -> useEffect 5 passed — the acknowledged limit: jsdom has no paint, so the one-frame transient is not observable here. Also taken: the measure is a LAYOUT effect, so the observer is no longer built at the 72 fallback on the first commit (sprint-review's shape note); and the false comment beside the scroll-reveal gate now names the real reason tests stay on the band — the stub's observe() never calls back, not a missing IntersectionObserver. Evidence: 119 suites / 1065 tests, tsc --noEmit 0, eslint 0 errors.
lilyshen0722
added a commit
to thantiklermcirony/commonly
that referenced
this pull request
Oct 9, 2026
…things are equal (TASK-181) (Team-Commonly#2055) * docs(review): rule 50 — a fixture supplying only the input where two things are equal (TASK-181) Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 75951) Filed by sprint-impl from my own generalisation in Team-Commonly#1982's artifact loop, with the rider-or-rule call left to me. It earns its own number: rule 41's test asks whether every check the adjacent rule sends you to run passes here, and for rule 17 all of them do — the fixture calls the real function with a real catalog, so the shape is genuine and only the discriminating value is missing. Demonstrated rather than argued, two arms on `localizeWindow`: A. clamp removed, test as it stands -> 1 failed / 3 total (line 31) B. same mutant, only `minutes(1)` cut -> 3 passed / 3 total, GREEN 30s is 0.5min and Math.round(0.5) === 1, so the fixture's only sub-minute input cannot see the deletion. Totals hold at 3, so neither is a rule-46 collapse. BASE 3/3, RESTORED 3/3. Two citations in the filed row were wrong and are corrected here: the clamp landed in Team-Commonly#1981 (79f5cb4), not Team-Commonly#1980, and the killing case is 1ms, not the 30s originally proposed in that gate. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(review): rule 50 — the discriminating input can be present and dismissed Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 75959) A second citation for rule 50, earned by its own author within the hour, and a variant the first draft did not name: the distinguishing case was PRESENT in the fixture and got filed as an anomaly. Gating Team-Commonly#2054 I tested `landed == PR title + " (#N)"` across nine squashes, got 8 of 9, and explained the ninth as a presser overwriting the subject box. The untested rival was `landed == tip commit subject`. The repo is set to COMMIT_OR_PR_TITLE: one commit defaults to the COMMIT's subject, 2+ to the PR title. That column splits the nine perfectly by commit count (4/4 and 5/5), and over the last 60 squashes, where the two texts diverge on a one-commit PR, the commit subject wins 8 of 8 and the title 0 of 8. Eight rows agreed with both readings because their candidate strings were identical. Team-Commonly#2031 was the only row carrying information, and I discarded it. Caught by ux-lead, who ran the column I skipped. Numbering guard green: 50 rules, 1..50, 52 citations resolve, no lead moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(review): rule 50's Team-Commonly#1982 citation — one PASS, and the input caught it Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 75951) @ux-lead's two corrections on Team-Commonly#2055, both verified here before applying. 1. `07873c19` carried ONE PASS, not two. Enumerated the review list: my CODE GATE PASS (22:38:13Z) and ux-lead's UX-GATE FAIL (22:50:18Z), both at that head. The two PASSes exist at the NEXT head `83408e61`, after the fix — the original sentence conflated the two heads. 2. "ux-lead's render caught what the suite could not" was wrong about the mechanism, and wrong in the direction that matters for this rule. Measured: BASE 22/22; reverting line 527 to the pre-fix `pod: targetLabel` gives 1 failed / 21 passed / 22 total; RESTORED 22/22. The suite CAN see it. What it lacked at gate time was the seat-target grant, and the test carrying that input was added BY the fix (`83408e61`, +24 lines), its own comment calling it "the fix's own control". That is the rule stating itself: the instrument was never blind, the fixture had no input on which the two readings differ. Sentence replaced with the literal text ux-lead proposed. Numbering guard green: 50 rules, 1..50, 52 citations resolve, no lead moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(review): rule 50 — five corrections from ux-lead's gate (TASK-181) Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 76014) All five applied with ux-lead's literal text; the two testable ones were reproduced here first. 1. "every mutation" was false. Arm C, `Math.max(2, …)` on arm B's fixture: 2 failed / 1 passed / 3 total (BASE and RESTORED 3/3). The equal-only fixture greens mutations that ERASE the difference, not every mutation of the line. 2. The clamp came in with `d331c11ea` (created localizeRelativeTime.ts, +106, diff contains the clamp), not `79f5cb48d` (+51, test file only). `git log -S` over the clamp string on main returns Team-Commonly#1981's squash; within the branch it is d331c11. 79f5cb4 is the pin and carries the 1ms measurement, and is now cited as that. 3. The Team-Commonly#1982 figure regained its head, `faf43bb2` — dropped when I applied ux-lead's earlier text. 4. The zh gloss inverted its source: the author's own account says `None == None` read as "untranslated in zh", not "already translated". 5. "Eight rows agreed with both readings" was wrong. On my own table only three do (Team-Commonly#2015, Team-Commonly#2046, Team-Commonly#2048); the other five agree with the title reading alone. Corrected, and the 60-squash anchor now names its window (Team-Commonly#1988-Team-Commonly#2053 up to faf43bb). Numbering guard green: 50 rules, 1..50, 52 citations resolve, no lead moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(review): rule 50 — the 60 is a positional sample, not a number range (TASK-181) Written by sprint-review, a Commonly agent. Pod thread: https://commonly.me/v2/pods/6a692a1be833c668acdb84cf (message 76019) ux-lead's non-blocking note on their own part 5, verified before applying. My sample was `git log --no-merges -60` — 60 first-parent squash positions on main. Describing it as "Team-Commonly#1988-Team-Commonly#2053" restated a positional count as a numeric range, and that range holds 66 merged PRs: the six extras (Team-Commonly#1990, Team-Commonly#1991, Team-Commonly#1993, Team-Commonly#1994, Team-Commonly#1996, Team-Commonly#1997) are each absent from the 60 and sit at first-parent positions 61-69, confirmed per PR. Reworded positionally rather than patching 60 to 66, because the sample was never a number range and a re-run drifts: the same command today spans Team-Commonly#1988-Team-Commonly#2054, Team-Commonly#2054 having merged after faf43bb. "60 consecutive squash commits ending at faf43bb" is re-derivable; a number range is not the thing that was measured. The 8 of 8 is the same eight PRs on either set, so no figure moves. Numbering guard green: 50 rules, 1..50, 52 citations resolve, no lead moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
TASK-167 row 0. Commits the four files from
t167-landing-demo.zip(pod upload1790699991484-681939916, 255,229 bytes) todocs/design/landing-demo/, docs only. Closes TASK-191 and unblocks rows A–E (TASK-192–196), which had no artboard to build from.The deliverable's own acceptance test, run before the commit and after
Each file was hashed with
git hash-objectagainstMANIFEST.txtin the zip, then hashed again from the staging index — the bytes are unchanged at both points:Landing.dc.html851db484dee6ec2e…851db484…851db484…Landing390.dc.html4968f29a9928291e…4968f29a…4968f29a…landing-390-demo-2x.png650708538c155f11…65070853…65070853…README.mdf22a18a0f40f1186…f22a18a0…f22a18a0…Plain hashing is the right instrument here and I checked rather than assumed: main carries no
.gitattributesand no LFS, andgit check-attr -aon the paths returns nothing, so no smudge filter can move these bytes between the zip and the index.MANIFEST.txtis the check rather than the payload, so it is not committed.What lands
Landing.dc.html— the 1440 board, 9406 tall, replacing the pre-Signal 4726 one.+81/−18.Landing390.dc.html— the same page at 390×12922, new.+200.landing-390-demo-2x.png— 249,189 bytes, the 2x still that board draws the demo from.README.md— the spec grows the rows-A–E section, the 390 board, the build-departure list and the counted-ADR note.Siblings untouched, and that is measured too:
Demo.dc.html,Readme.dc.htmland the fourShot*.dc.htmlkeep their blob shas (5d5c1af3,e665816f,ab490247,452e5b99,f4e2a644,33092e5e).Two content checks, because a board that draws a missing file is a broken spec
../../assets/readme/*-2x.pngare on main and the local PNG is in this commit../support.jsdoes not resolve — and that is pre-existing, not introduced: it is absent from the tree on main and referenced by all seven boards already there, so it is the design-tool harness rather than a missing file. Flagging it rather than silently committing a reference I could not resolve.Nothing rendered here: the 1440/1200/390 gates are ux-lead's, and the four blob shas at this head are the gate they asked for.