Re: [RFT][PATCH v3] gcc: upgrade 11.2 -> current gcc-12 snapshot

Khem Raj

On Sat, Apr 16, 2022 at 1:07 AM Richard Purdie
<richard.purdie@...> wrote:

On Fri, 2022-04-15 at 10:38 -0700, Khem Raj wrote:
From: Bernhard Rosenkränzer <bernhard.rosenkraenzer.ext@...>

- gcc 12 is in RC stages.
- Add patch to re-shuffle include of sched.h to fix build on musl

Signed-off-by: Bernhard Rosenkränzer <bernhard.rosenkraenzer.ext@...>
Signed-off-by: Khem Raj <raj.khem@...>
- Switch to own snapshot for quick turnarounds than waiting for a
weekly snapshot
- Add a fix for gcc target build on musl
- Fixed gcc for kernel build failures on arm/x86_64
- Fixed gcc for ccache build failure
I ran this through the autobuilder as requested as it means we have a chance to
get things fixed upstream before gcc 12 is released:

We see failures in:

edgerouter/edgerouter-alt in linux-yocto 5.15/5.10:
these are fixed with
and I see its already backported to 5.15 branch. so whenever Bruce
bumps to latest tip of linux-5.15 we should be
getting this fixed.

meta-arm in linux-yocto 5.15:
this can be addressed by disabling -Werror=stringop-overread but
perhaps better to change code to fix it.

meta-intel in intel-media-driver:
error: the address of '_CODEC_VP8_PIC_PARAMS::ucMvUpdateProb' will
never be NULL [-Werror=address]
| 238 | if (codecPicParams->ucMvUpdateProb[0] && picParam->mv_probs[0])

perhaps fixable by not checking the redundant
qemuppc/qemuppc-alt in linux-yocto 5.15/5.10:
include/linux/printk.h:450:44: error: using a dangling pointer to
'__str' [-Werror=dangling-pointer=]
this needs fixing too. I see it on ppc64 as well.

Four oe-selftest failures due to missing maintainer entries (easy fix)
I will add it in v4.

A weird oe-selftest prservice.BitbakePrTests.test_import_export_override_db:

The latter is one of our infamous intermittent issues and likely not gcc 12
Parsing recipes...ERROR: ParseError in None: Not all recipes parsed, parser
thread killed/died? Exiting.
ERROR: Parsing halted due to errors, see error messages above

So basically you have the intel-media-driver failure in meta-intel and a number
of kernel issues. I've cc'd a few of the maintainers just so they're aware of
the status, I appreciate we'd wait for the mainline kernel to resolve things.



Join to automatically receive all group messages.