[OE-core] M3 build status

[Richard Purdie]

On Sun, 2020-03-08 at 13:18 +0000, Richard Purdie wrote:
> On Sun, 2020-03-08 at 14:08 +0100, Alexander Kanavin wrote:
> > On Sun, 8 Mar 2020 at 09:15, Richard Purdie <
> > richard.purdie at linuxfoundation.org> wrote:
> > > This is now closer. Alex fixed the libmodule-build-perl and
> > > gettext
> > > reproducibility issues (thanks!), I tracked down a glibc issue
> > > and
> > > procps one I saw locally. On a fresh autobuilder run we saw:
> > > 
> > > https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20200307-mk0v6ljq/packages/diff-html/
> > > and
> > > https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20200307-5pvgdd3a/packages/diff-html/
> > > 
> > > which I believe is the remaining blocking issue on coreutils-
> > > ptest.
> > 
> > Not quite. The coreutils-ptest itself seems to persistently fail
> > here:
> > 
> > AssertionError: Failed ptests:
> > {'coreutils': ['tests/tail-2/assert.sh']}
> > 
> > so I'd like to have that fixed before it goes in. We have a 100%
> > ptest pass rate now (except for random timing fluctuations), which
> > I
> > don't want to see regressed :)
> 
> I agree that would be nice and something we need to do before
> release,
> I may let us sort that one test in M4 to enable the M3 build, I have
> to
> take a pragmatic approach to this.
> 
> That said, there is a more pressing issue:
> 
> https://autobuilder.yoctoproject.org/typhoon/#/builders/44/builds/1686
> 
> which is somehow seemingly related to the coreutils change, I just
> can't see how as yet :/
> 
> (as well as the gcc-plugin reproducibility issues)

Caused by the new libmodule-build-perl which depends upon build-
essential and hence gcc which isn't target multilib 'safe'.

If we could stop the -ptest package dependencies leaking to -dev we'd
avoid this.

Cheers,

Richard