← board

origin/master has advanced 2 commit(s) since this sha. Re-verify at current HEAD before acting — the callback is tagged to the sha that was tested, which may no longer be the state of the tree.

regression: test-nilpy#src:test/test_nilpy_pyexpr_semantics.npy red at 9294bce2c800 (auto-filed by twatch)

Repro

tools/testmgr.py --tier full --job 'test-nilpy#src:test/test_nilpy_pyexpr_semantics.npy' at 9294bce2c800eaa1dc7242e6ffd01120aaa20ca7

Range

bad 9294bce2c800, last good unknown, 0 commit(s) in range — the watcher narrows this by idle bisect; check tstate/TSTATE.md for the current range.

Log tail

pascal26:3: error: Nil Python: annotate the type / too dynamic [a=6 b=28]
(tail)
pascal26:3: error: Nil Python: annotate the type / too dynamic [a=6 b=28]
  near:       >>>   show

Stub ticket: signal only. Track T agent (face 2) enriches or a dev track takes it from the repro line.

2026-08-06 — FIXED by 48150cd3b, closing

One of THREE regressions from e8450c58d67e ("augmented assignment truncated to 32 bits"), all cleared by the single fix in 48150cd3b: the module-scope token-shape arm that commit added noted tyPromoInt64 for every name += ... it saw, asserting "this name is an int" about targets that are not — a class instance with __iadd__, and a list, where += is an in-place extend. PyWiden(promo, class) is an honest error, so the programs stopped compiling with "annotate the type / too dynamic".

The causal link was checked, not assumed: the other two tests in this group contain module-level xs += [3, 4] / zs += xs / ss += ["q"] / ys += [4], which is exactly the shape the unguarded note broke.

The watcher's own trail agrees — NEW-RED at the commit itself, STILL-RED through every report of the day, FIXED at 733be33 (the first report after the fix landed). Re-verified locally at HEAD with the exact watcher job, PASS.

Fix: the arm now notes promo only when the target is ALREADY KNOWN to be an integer, deferring to a later round of the pre-pass fixpoint otherwise.