[OE-core] [PATCH] rpm: fix build issue with gcc5

[Richard Purdie]

On Sun, 2015-05-10 at 23:03 +0100, Richard Purdie wrote:
> On Sun, 2015-05-10 at 09:15 -0700, Khem Raj wrote:
> > On Sun, May 10, 2015 at 7:28 AM, Armin Kuster <akuster808 at gmail.com> wrote:
> > > | tset.o:(.debug_loc+0xe8): undefined reference to `.LCL0'
> > > | tset.o:(.debug_loc+0xf7): undefined reference to `.LCL0'
> > >
> > > thanks Khem
> > >
> > > Signed-off-by: Armin Kuster <akuster808 at gmail.com>
> > > ---
> > >  meta/recipes-devtools/rpm/rpm_5.4.14.bb | 3 +++
> > >  1 file changed, 3 insertions(+)
> > >
> > > diff --git a/meta/recipes-devtools/rpm/rpm_5.4.14.bb b/meta/recipes-devtools/rpm/rpm_5.4.14.bb
> > > index 03a24f0..22b111a 100644
> > > --- a/meta/recipes-devtools/rpm/rpm_5.4.14.bb
> > > +++ b/meta/recipes-devtools/rpm/rpm_5.4.14.bb
> > > @@ -110,6 +110,9 @@ inherit autotools gettext
> > >
> > >  acpaths = "-I ${S}/db/dist/aclocal -I ${S}/db/dist/aclocal_java"
> > >
> > > +FULL_OPTIMIZATION_powerpc_remove = "-O2"
> > > +FULL_OPTIMIZATION_append_powerpc = " -Os "
> > > +
> > 
> > http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/master&id=741d3bf91144973d26de647039fa60b87a09edf1
> > fixes same issue for libx11
> > 
> > I am wondering if we should make this change in some common config
> > metadata file to list all recipes that need this workaround
> > untill gcc is fixed for it may be arch-powerpc.inc, since then it will
> > be easy to remember to unbolt them once gcc is fixed.
> 
> I like the idea of a common include to handle this a lot better.
> 
> FWIW, I tried gcc 5 on the autobuilder. The results were 141 failures so
> far:
> 
> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150

With the gcc patch I've posted applied:

http://errors.yoctoproject.org/Errors/Search/?items=10&query=5fd0b7c163f4f3312bea9bd3246a6bd67a8da594&limit=100

so down to 57 errors. There are a few themes:

* linux-yocto 3.14 gcc5 issues
* meta-fsl-ppc and metafsl-arm gcc5 issues, particularly kernel
* poky-lsb issues with the security flags and gcc5
* x32 failed with a race of some kind in glibc, suspect transient
* gcc-target has a packaging issue which fails builds (have patch)

The good news is that the various ppc issues are fixed including the rpm
one from this series.

Ross: The backports we were looking at are for warnings, not errors on
the most part.

Otavio: Could you see what our options are with meta-fsl-ppc and
metafsl-arm?

Cheers,

Richard