[poky] glibc-initial fails to build in Thud with arm926ejs tune

[Matthew Stanger]

>
> Can give follow on config.log's and such as required.

Yep, I'm using the default 8.2.0 default.

The mcpu verses mtune is open to debate, it depends what your priority
> is, feed compatibility across multiple processors or potential micro
> optimisation for the processor.

This machine.conf/meta layer just needs to support a single target - micro
optimization is definitely desired since it's a potato ARM. I'm guessing I
could just override the 'TUNE_CCARGS' in my machine.conf to '-mtune' then?

Thanks


On Wed, Jan 2, 2019 at 2:17 PM Richard Purdie <
richard.purdie at linuxfoundation.org> wrote:

> On Wed, 2019-01-02 at 13:37 -0700, Matthew Stanger wrote:
> > I'm getting a build failure under Thud when trying to use the
> > following tune in my machine.conf:
> > #at91sam9.inc is from meta-atmel - at91sam9263 is ARM target
> > require conf/machine/include/at91sam9.inc
> > DEFAULTTUNE = "arm926ejs"
> >
> >
> > When doing a fresh build 'glibc-initial' config stage fail's with:
> > > checking for broken __attribute__((alias()))... yes
> > > configure: error: working alias attribute support required
> > > WARNING: exit code 1 from a shell command.
> > > ERROR: Function failed: do_configure (log file is located at
> > /home/matt/renew_thud/build/tmp/work/arm926ejse-poky-linux-
> > gnueabi/glibc-initial/2.28-r0/temp/log.do_configure.1177)
> > ERROR: Task (/home/matt/renew_thud/build/../oe/poky/meta/recipes-
> > core/glibc/glibc-initial_2.28.bb:do_configure)
> > failed with exit code '1'
> > NOTE: Tasks Summary: Attempted 2173 tasks of which 0 didn't need to
> > be
> > rerun and 1 failed.
> >
> > Summary: 1 task failed:
> >   /home/matt/renew_thud/build/../oe/poky/meta/recipes-
> > core/glibc/glibc-initial_2.28.bb:do_configure
> >
> > Can give follow on config.log's and such as required.
> >
> > If I edit 'poky/meta/conf/machine/include/tune-arm926ejs.inc' to
> > change the '-mcpu' back to 'mtune'(commit f29d642c) then glibc builds
> > fine. GCC tuning is totally outside my expertise so I'm hoping
> > someone
> > can confirm if '-mcpu' is really desired over '-mtune'? And also if
> > this look like a Poky issue or something else?
>
> Are you using thud's default gcc version? I have some memory this issue
> may be gcc version specific.
>
> The mcpu verses mtune is open to debate, it depends what your priority
> is, feed compatibility across multiple processors or potential micro
> optimisation for the processor.
>
> Cheers,
>
> Richard
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/poky/attachments/20190102/b8b6af73/attachment.html>