Repository navigation
tests/section_weights.txt is stale: 46 sections are unmeasured and take the 0.5 s default, including #1776's [0fk] (~53 s) #1784
Description
Found by the #1758 R1.1 blind critic (PR #1776), 2026-10-14.
bash tools/section_plan.sh --shards 3 --check on 9ce8e0b prints:
SHARDS: n=3 chunks=490 groups=476 (merges=14) union=full disjoint=yes zero-section-shards=0 unmeasured-sections=46
The weights were measured at run 36102915685 (head 05bda90). Every section added since takes SHARD_DEFAULT_CS, 0.5 s. One of them is #1776's [0fk]: it took 53.33 s on the PR's linux / gcc lane, the second slowest of 283 sections, and it compiles the amalgamation at -O2 in the ASan shard as well. The planner can therefore place about a minute of unmeasured work on the heaviest shard. In merge-queue run 38652383231 the three ASan shards already took 10.0, 11.1 and 9.8 minutes.
Done when
-
tests/section_weights.txtis refreshed from the shard logs of one CI run (the recipe in its header), andbash tools/section_plan.sh --shards 3 --checkprintsunmeasured-sections=0at that head. - Once Phase 2 numbers: exact int64 and IEEE doubles (#1758) #1776 has merged, [0fk] is in the file with its measured weight.
Activity
phase2 R1.1 fix: the slot test's full matrix, its channels and its va…
InauguralPhysicist commentedon Oct 15, 2026
From the #1758 R1.1 round-2 blind critic (2026-10-15), on the fix in #1776 (d807683):
The
[0fk] 53.33row intests/section_weights.txtcomes from thelinux / gcclane at 9ce8e0b. There the slot test compiled the release flavour at -O2. The planner places that weight on the ASan shards, and after #1783's fix (same commit) those shards build the ASan variant instead (slot_layer.sh asan, the amalgamation at -O1 with ASan and UBSan). On the 4 GB dev box that variant takes 4m34s (time bash tests/phase2/slot_layer.sh asan, cdfb128), against 45.56 s for the release[0fk]in a local suite run.section_plan.sh --shards 3 --checkprintsunmeasured-sections=0, but this one weight was measured for a different build than the one the shards run. The first CI ASan shard log after #1776 runs CI again should replace it.