On Mon, 16 Feb 2026 at 20:48, Ankur Tyagi via lists.openembedded.org
<[email protected]> wrote:
> > > Trying to understand how come do_install() succeeded but
> > > https://valkyrie.yocto.io/pub/repro-fail/oe-reproducible-20260214-8ytqayuf/packages/diff-html/
> > > show sed substitution failure.
> > > Can you please help point me towards the mdadm recipe logs?
> > The problem is I probably won't have the log that you are really
> > interested in: left side (A) is built using sstate-cache, only the right
> > side (B) is built locally.
> >
> > So we do not have the log of the build where the sed failed, as it comes
> > from the cache. We only have the log for the build where it succeed. I'm
> > still joining do_install logs to the mail.
> >
> > What I can do is run a few other builds, hopping that at some point the
> > sed error will happen during B build.
> I have made some improvements in the Makefile install rule and will
> soon send v2 for the test.
I have this horrible suspicion that the BINDIR item still comes from
sstate produced by a previous build via a bogus hash equivalence
entry, which is why the error never triggered. Which means, when you
send a new patch, make sure to also install something harmless into
${D} from do_install, so that the recipe output is different from any
previous build.
Alex
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#231223):
https://lists.openembedded.org/g/openembedded-core/message/231223
Mute This Topic: https://lists.openembedded.org/mt/117786134/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
