Repository navigation
t27b lane 4: @abs, @max, @min and std.math.pi / std.math.e #7394
Description
Part of #6063 (t27b lane 4).
Three blocked specs stop at Zig builtins and a std.math constant that the reference prints as written and t27b refuses:
specs/ml/rl/dqn.t27:ExprCall(@max)at line 82 (td_target), thenExprCall(@abs)in its tests. Its tests became real in specs: dqn, hex, gen_image, args rewritten from their sources with real tests (Closes #7372) #7375.specs/port/trinity/src/tri/string_dualities.t27:ExprFieldAccess(std.math.pi), thenExprCall(@abs).
The plan is one backend slice in cli/t27b:
@abs(x)lowers the way a bareabs(x)already does, because the reference prints both as Zig's@abs. A typed integer operand is still refused, because Zig's@absof aniNis auN.@max(a, b)and@min(a, b)take two operands. Zig 0.16 in Debug was measured on the t27b lab:- Floats of one type: a NaN operand gives the other operand. When the operands compare equal (
-0.0,+0.0), the first operand is the result. - A literal beside a typed float takes that float's type.
- Integers of one type give the larger or smaller value.
- Each operand is evaluated once.
- Floats of one type: a NaN operand gives the other operand. When the operands compare equal (
- Still refused:
- an integer beside a literal (Zig narrows the result type);
- mixed types;
- two literals;
- more than two operands.
std.math.piandstd.math.ebecome the comptime_float literals that Zig'sstd/math.zigdeclares, rounded to binary128 as Zig rounds them. The otherstd.mathnames stay refused.
The conformance fixture specs/tri/t27b/conformance/builtin_abs_max_min_pi.t27 comes first. It passes t27c test-report 12/12 with 0 vacuous tests before any Rust changes.
Boundary
specs/ml/rl/dqn.t27specs/port/trinity/src/tri/string_dualities.t27specs/tri/t27b/conformance/builtin_abs_max_min_pi.t27
Activity
t27b: @abs, @max, @min and std.math.pi / e lower; dqn and string_dual…
gHashTag commentedon Oct 7, 2026
Foreign-line budget (owner rule #7371, merged in #7399): PR #7411 is over it, so it is now a draft with auto-merge off.
Master's
check_budget()run ongit diff --numstat --no-renames origin/master...claude/t27b-lane4-abs-max-pi:cli/t27b/src/lower_float.rscli/t27b/src/lower.rsThe
owner-approved-foreignlabel does not lift this cap, and splitting the work into smaller PRs to get under it would dodge the rule, so I am not doing that. The lowering of@abs/@max/@min/std.math.pihas to move into.t27and come out of gen-rust, with only thin glue left by hand. The conformance specspecs/tri/t27b/conformance/builtin_abs_max_min_pi.t27and the ledger rows (dqn, string_dualities) stay valid for that rework.Part of #6063. PR: #7411.
gHashTag commentedon Oct 7, 2026
Reopened 2026-10-07: the bulk close of issues without a
## Boundarysection was reverted by the owner. Nothing to do here; this comment replaces the earlier close notice.t27b ledger: six specs move to pass, the two new specs are listed (Cl…
t27b: builtin_plan.t27 tests compare strings with std.mem.eql (Closes #…
t27b: no conformance test pins the sign of @max / @min of -0.0 and +0…