On Wed, 2024-08-14 at 16:39 -0700, Tim Orling via lists.openembedded.org wrote: > > > On Wed, Aug 14, 2024 at 3:51 PM Alexandre Belloni > <[email protected]> wrote: > > This causes reproducibility issues: > > > > > lol. This was supposed to fix reproducibility issues! > > > > > https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20240813-qgfyyhkg/packages/diff-html/ > > > > > > > Those data.tar.xz differences look entirely different. I'm going to > have to dig over the next couple days and perhaps ask upstream, > unless someone else recognizes a pattern. Although I wonder if this > is just "this rust package was not populated in Repro A because the > task was already in state and it was populated in Repro B because it > was built from scratch". If that is the case, we have something > elsewhere causing problems with state (rust, cargo, python3-py03 or > python3-maturin for instance). >
They don't look that different. The RECORD file changed and the lines look to have changed ordering as I think the same lines are there, e.g.: cryptography/hazmat/primitives/hashes.py,sha256=EvDIJBhj83Z7f- oHbsA0TzZLFSDV_Yv8hQRdM4o8FD0,5091 and cryptography/hazmat/primitives/hashes.py,sha256=EvDIJBhj83Z7f- oHbsA0TzZLFSDV_Yv8hQRdM4o8FD0,5091 but different ordering. So the question is what is writing that file and is it doing it based on something like "ls" output which just lists the order on disk? If that is true you can thrown a sort in somewhere... Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#203336): https://lists.openembedded.org/g/openembedded-core/message/203336 Mute This Topic: https://lists.openembedded.org/mt/107870272/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
