Le lun. 14 août 2023, 10:46, Richard Purdie <
[email protected]> a écrit :

> On Mon, 2023-08-14 at 09:30 +0200, Frédéric Martinsons wrote:
> >
> >
> > On Sun, 13 Aug 2023 at 17:05, Richard Purdie
> > <[email protected]> wrote:
> > > On Sun, 2023-08-13 at 17:00 +0200, Frédéric Martinsons wrote:
> > > > On Sun, 13 Aug 2023 at 16:53, Richard Purdie
> > > > <[email protected]> wrote:
> > > > >
> > > > > and a reproducibility failure:
> > > > >
> > > > >
> https://autobuilder.yoctoproject.org/typhoon/#/builders/117/builds/3355/steps/13/logs/stdio
> > > > >
> > > > > which leads to:
> > > > >
> > > > >
> http://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20230813-z_b2j3ha/packages/diff-html/
> > > > >
> > > >
> > > >
> > > > Argh, this makes me remember
> > > > of https://bugzilla.yoctoproject.org/show_bug.cgi?id=15090
> > > > Do you know if any of cargo based recipe is reproducible ?
> > > > Should I add EXCLUDE_FROM_WORLD in cargo-c ?
> > >
> > > At some point we're going to have to dive in and fix the
> > > reproducibility issues so I'm reluctant to take more patches with
> > > that
> > > set...
> > >
> >
> >
> > It will hardly be doable to correct the reproducibility of rust
> > packages since rust/cargo
> > itself have issues upstream. I take a look and there are lot of open
> > issues
> > cargo:
> > https://github.com/rust-lang/cargo/issues?q=is%3Aissue+is%3Aopen+labe
> > l%3AA-reproducibility+
> > rust:
> > https://github.com/rust-lang/rust/issues?q=is%3Aissue+is%3Aopen+label
> > %3AA-reproducibility
>
> I believe there is one problematic reference left which causes problems
> for core rust. Fixing that issue would be the best place to start and
> then we can work from there.
>
> We do already have a lot of good reproducibility pieces in the
> toolchain so not every upstream bug will apply to us.
>

I'm afraid I do not see what is your point, can you elaborate please? What
is the "problematic reference left which causes problems
for core rust" you refer to?


> > There was an RFC accepted back in may
> > (https://github.com/rust-lang/rfcs/pull/3127) and implementation
> > progress have a tracking issue in cargo
> > (https://github.com/rust-lang/cargo/issues/12137) and in rust
> > (https://github.com/rust-lang/rust/issues/111540).
> >
> > Don't know if it will fix our root issue
> > (https://bugzilla.yoctoproject.org/show_bug.cgi?id=14875) but I don't
> > see how to workaround all of that without this upstream work.
> >
> > For cargo-c specific issues, I'll try to make things better (the
> > TMPDIR in debug symbols) but I'm wondering
> > something, reproducibility is  desirable for all generated artifacts
> > or is it only for target ones ?  I mean,
> > the cargo-c is used in native so I'm wondering if making it native
> > only would be acceptable ?
>
> That would certainly avoid this issue. It would also make on target
> testing of the functionality not possible.
>

The test I added is on the final binary generated (rust-c-lib-examble-bin),
not cargo-c itself. So the target test would still be valid.

I tested a release build with a custom patch that strips debuginfo
(strip="debug info" in Cargo.toml) and I do not have build path leaking in
debug symbol anymore. I'll dig this up more and see if I can upstream that.

But the reproducible issue remain despite that.


> Cheers,
>
> Richard
>
>
>
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#185928): 
https://lists.openembedded.org/g/openembedded-core/message/185928
Mute This Topic: https://lists.openembedded.org/mt/100715215/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to