From f28cac06912bdc1719ee41bb0b61a4feab38af4a Mon Sep 17 00:00:00 2001 From: Jack Champagne Date: Mon, 8 Jun 2026 15:10:47 -0400 Subject: [PATCH] ci(alloc): bump alloc-profile timeout 90 -> 120 min for Julia 1.12 The allocation profile runs each solve under Profile.Allocs sampling (~30-40 min per solve). On Julia 1.11 the full run was ~76 min; on 1.12 it overran the 90 min cap (cancelled at 91 min), so /bench-alloc never seeded. Bump to 120 min to give the slower 1.12 runtime real headroom. Co-Authored-By: Claude Opus 4.8 (1M context) --- .github/workflows/alloc-profile.yml | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/.github/workflows/alloc-profile.yml b/.github/workflows/alloc-profile.yml index e2ae654..8f7791f 100644 --- a/.github/workflows/alloc-profile.yml +++ b/.github/workflows/alloc-profile.yml @@ -21,8 +21,9 @@ jobs: # Profile.Allocs has high per-allocation overhead that doesn't scale down # with sample_rate — each Ipopt/MadNLP solve under sampling takes ~30-40 # min on GH Actions runners even at max_iter=30. Two solves + startup = - # ~75 min observed in practice; 90 min gives a comfortable cushion. - timeout-minutes: 90 + # ~75 min on Julia 1.11, but ~90+ min on Julia 1.12 (it overran a 90 min + # cap), so 120 min gives a real cushion on the slower runtime. + timeout-minutes: 120 permissions: actions: write # contents: write so github-action-benchmark can auto-push the time-series