Hi Sunil, On Thu, 2026-09-10 at 13:25 +0000, Dora, Sunil Kumar via lists.openembedded.org wrote: > On 9/9/2026 4:02 PM, Richard Purdie via lists.openembedded.org wrote: > > d) Reproducibility on x86 vs arm hosts. Alejandro has made good > > progress with identifying many of the rust related issues but there > > are > > some non-rust issues remaining. One key set of diffs are these: > > > > https://valkyrie.yocto.io/pub/repro-fail/oe-reproducible-20260908-sv8uww77/packages/diff-html/ > > > > Particularly the perl issue, the locale change and numpy config > > issues. > > > > Any help on these welcome, just let us know if you're working on > > anything to avoid duplication. > > On point (d), just to avoid duplicating work: I’ve been looking into > the x86 vs. ARM host reproducibility failures for some time now, > mainly around the Rust and Git issues. > For rust/rust-dbg, the remaining diff after the SVH work appears to > be caused by the build path leaking into object file names > during the Rust build. I have a fairly good understanding of the > issue now and am validating a fix locally. > Once confirmed, I’ll post the analysis and patch. > I’ve also been looking into the Rust test suite failures seen with > Alejandro’s v5 series, and I believe I understand what’s happening > there. > I’ll follow up separately on that thread with more details and a > suggested rework, so the SVH improvements aren’t lost. > For git/git-dbg, this is theHOST_CPU issue I mentioned earlier in the > thread. I’ve since confirmed the fix locally and will send the patch > shortly. > For git/git-dbg, this is theHOST_CPU issue I mentioned earlier in the > thread. I’ve since confirmed the fix locally and will send the patch > shortly. > I’m not currently looking into the Perl, Locale, or NumPy diffs, so > those are available for anyone else to pick up.
Thanks, this is good information to have and sounds like really good progress. I'm looking forward to getting this resolved! I had remembered about the git issue which is why I didn't call that one out specifically. Ross has sent a patch for numpy too, so we are moving forwards. The other reproducibile tests did complete and the results there are looking better too, the rust "noise" is reduced showing the other issues more clearly: https://autobuilder.yoctoproject.org/valkyrie/#/builders/121/builds/4 https://valkyrie.yocto.io/pub/repro-fail/oe-reproducible-20260908-nxt738yt/packages/diff-html/ https://autobuilder.yoctoproject.org/valkyrie/#/builders/119/builds/4/steps/15/logs/stdio https://valkyrie.yocto.io/pub/repro-fail/oe-reproducible-20260908-xz_trxrt /packages/diff-html/ There are kernel, gcc, go, mesa, pseudo, erofs and ruby issues there so plenty for people to help with if anyone has time! Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#245599): https://lists.openembedded.org/g/openembedded-core/message/245599 Mute This Topic: https://lists.openembedded.org/mt/121159956/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
