From eb51a8a194d73738aaac274c48f2cfa95efc2085 Mon Sep 17 00:00:00 2001 From: ScriptedAlchemy Date: Tue, 25 Aug 2026 01:04:53 +0000 Subject: [PATCH] ci(windows): raise build timeout for cold cache `Build Windows tests` no longer fails to compile -- 727a6d6c3 fixed the E0425 that broke the Windows lib-test unit -- but the job now dies at the 45m job timeout instead. In run 32789438873 the Swatinem cache restore took 12s (a miss), setup finished in 52s, and `Build nextest archive` then burned 44m08s before the job timeout killed it mid-compile. Earlier runs only appeared to fit inside 45m because the resolution error aborted the `tracedecay` lib-test unit before codegen. Raise the ceiling to 75m so a genuinely cold cache can complete. The gate still runs and still has to pass; a hung linker will fail loudly. Co-Authored-By: Claude Opus 5 (1M context) Claude-Session: https://claude.ai/code/session_0175M5Lyn1pPrX9a6CYvwDVn --- .github/workflows/ci.yml | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index b50fa55da..61284d9aa 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -257,9 +257,13 @@ jobs: if: ${{ needs.scope-gate.outputs.run-heavy == 'true' }} runs-on: windows-latest # Full-workspace `nextest archive` on hosted windows-latest has been - # dying at exactly 30m with no failed step. 45 leaves headroom for a - # cold cache without hiding a hung linker. - timeout-minutes: 45 + # dying at the job timeout with no failed step: first at 30m, then at 45m + # (run 32789438873 burned 44m08s in `Build nextest archive` on a cold + # Swatinem cache -- the restore step took 12s, i.e. a miss -- and was + # killed mid-compile). Earlier runs only "finished" faster because a + # resolution error aborted the `tracedecay` lib-test unit early. 75m fits + # a genuinely cold cache while still failing loudly on a hung linker. + timeout-minutes: 75 steps: - uses: actions/checkout@v7