Bug 2226990
| Summary: | Firefox build failure: error: could not compile `gkrust` (lib), (signal: 9, SIGKILL: kill) | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Martin Stransky <stransky> |
| Component: | rust | Assignee: | Rust SIG <rust-sig> |
| Status: | NEW --- | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
| Severity: | medium | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 38 | CC: | amulhern, decathorpe, igor.raits, jistone, rust-sig, TicoTimo |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | Unspecified | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | If docs needed, set a value | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | Type: | --- | |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Martin Stransky
2023-07-27 10:06:59 UTC
This looks like rustc is getting reaped by the OOM killer ... Can you check whether your build on F37 used Rust 1.71 or still 1.70? If it's the latter, then this looks like a new problem with Rust 1.71. Looks like so. Latest successful Firefox 115.0.2 build for Fedora 38 was done with rustc version 1.70.0 on Thu, 20 Jul: https://kojipkgs.fedoraproject.org//packages/firefox/115.0.2/3.fc38/data/logs/x86_64/build.log But recent builds with rustc version 1.71.0 fails (build config is almost the same). I see you now have a successful build on f38 with 1.71.0 -- what changed? https://koji.fedoraproject.org/koji/buildinfo?buildID=2264243 (In reply to Josh Stone from comment #3) > I see you now have a successful build on f38 with 1.71.0 -- what changed? > https://koji.fedoraproject.org/koji/buildinfo?buildID=2264243 Nothing significant - just changelog update and rawhide python build fix. If the failure is caused by OOM maybe different build box with more ram was chosen by koji? Hmm, yes, the failed build was on buildvm-x86-07 with 16 GiB, and it succeeded on buildhw-x86-02 with 128 GiB. Maybe you can reduce parallelism when there's less memory available? Or invoke the gkrust build separately, before the full build? (Assuming that killed process didn't consume all memory on its own -- LTO might do that! But it looks like other C++ processes were active too.) Since you're already using the %constrain_build macro for firefox: %constrain_build -m 2048 It might be worth trying to set this to a higher value? %constrain_build -m 4000 seems to be working (somehow). |