CI: Intermittent "Command failed with error code: -4" on Darwin.
Recent occurrences:
| json->>'web_url' | json->>'status' | json->>'ref' | json->>'name' | json->>'runner.description' | date(json->>'created_at') |
|---|---|---|---|---|---|
| https://gitlab.haskell.org/ghc/ghc/-/jobs/1772620 | failed | refs/merge-requests/11968/head | x86_64-darwin-validate | m1-delta_Mini18-Delta_f87b8a204711 | 2024-02-07 |
| https://gitlab.haskell.org/ghc/ghc/-/jobs/1767573 | failed | refs/merge-requests/11576/head | x86_64-darwin-validate | aarch64-darwin-1_Mini18-Beta_f87b8a204711 | 2024-02-01 |
| https://gitlab.haskell.org/ghc/ghc/-/jobs/1767305 | failed | refs/merge-requests/11976/head | x86_64-darwin-validate | aarch64-darwin-1_Mini18-Epsilon_004e0928d81c | 2024-02-01 |
| https://gitlab.haskell.org/ghc/ghc/-/jobs/1764568 | failed | refs/merge-requests/11989/head | x86_64-darwin-validate | aarch64-darwin-1_Mini18-Beta_f87b8a204711 | 2024-01-29 |
| https://gitlab.haskell.org/ghc/ghc/-/jobs/1753166 | failed | refs/merge-requests/11923/head | x86_64-darwin-validate | aarch64-darwin-1_Mini18-Beta_f87b8a204711 | 2024-01-16 |
Analysis
- Signal 4 is "Illegal instruction", so maybe Rosetta is involved.
- The (truncated) command lines that failed in these occurences are:
/nix/store/sbfhs3rv53vgh3cn46q900fmgsl5q4x4-ghc/bin/ghc-pkg --no-user-package-db field filepath depends_build/stage0/bin/ghc -Wall -Wcompat -fdiagnostics-color=never -dynamic-too -hisuf hi -osuf o -hcsuf <SNIP>_build/stage0/bin/ghc -Wall -Wcompat -hisuf dyn_hi -osuf dyn_o -hcsuf dyn_hc -fPIC -dynamic <SNIP>/nix/store/sbfhs3rv53vgh3cn46q900fmgsl5q4x4-ghc/bin/ghc-pkg --expand-pkgroot --no-user-package-db describe stm/usr/bin/gcc --target=x86_64-apple-darwin -Qunused-arguments -E -MM -MG -MF _build/stage1/rts/build/c/sm/NonMovingScav.thr_debug_o.d <SNIP>
So if it managed to affect nix-packaged software, our own built software, and /usr/bin/gcc, I feel it has to be a system issue. Unfortunately it also seems to affect multiple machines.
Tracking link
Edited by Bryan R