-D_GNU_SOURCE kselftest breakage in mainline

-D_GNU_SOURCE kselftest breakage in mainline

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 704 bytes --]

Hi,

I'm seeing quite a lot of breakage in mainline as a result of
daef47b89efd0b7 ("selftests: Compile kselftest headers with
-D_GNU_SOURCE") and daef47b89efd0 ("selftests: Compile kselftest headers
with -D_GNU_SOURCE") - thus far I've found that the use of
static_assert() is triggering build breaks where testsuites aren't
picking up the addition of _GNU_SOURCE (including stopping installing
the other tests in the same directory), and there's a bunch of tests
which #define _GNU_SOURCE in their code and now trigger build warnings.
I'm looking at fixes and mitigations now.

The build failures are taking out the ALSA tests entirely which has
caused my personal CI to explode badly :/

Thanks,
Mark

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: -D_GNU_SOURCE kselftest breakage in mainline

[Shuah Khan]

On 5/16/24 08:02, Mark Brown wrote:
> Hi,
> 
> I'm seeing quite a lot of breakage in mainline as a result of
> daef47b89efd0b7 ("selftests: Compile kselftest headers with
> -D_GNU_SOURCE") and daef47b89efd0 ("selftests: Compile kselftest headers
> with -D_GNU_SOURCE") - thus far I've found that the use of
> static_assert() is triggering build breaks where testsuites aren't
> picking up the addition of _GNU_SOURCE (including stopping installing
> the other tests in the same directory), and there's a bunch of tests
> which #define _GNU_SOURCE in their code and now trigger build warnings.
> I'm looking at fixes and mitigations now.
> 

Would it be better to revert this for now and get this for now? I wouldn't
want you to extra busy work to workaround this.

> The build failures are taking out the ALSA tests entirely which has
> caused my personal CI to explode badly :/
> 

This has been in next for a while and I didn't see any failures. These
kind of changes are tricky. On second thought, I probably should have
delayed picking this up.

thanks,
-- Shuah

Re: -D_GNU_SOURCE kselftest breakage in mainline

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 1764 bytes --]

On Thu, May 16, 2024 at 08:53:52AM -0600, Shuah Khan wrote:
> On 5/16/24 08:02, Mark Brown wrote:

> > I'm seeing quite a lot of breakage in mainline as a result of
> > daef47b89efd0b7 ("selftests: Compile kselftest headers with
> > -D_GNU_SOURCE") and daef47b89efd0 ("selftests: Compile kselftest headers
> > with -D_GNU_SOURCE") - thus far I've found that the use of
> > static_assert() is triggering build breaks where testsuites aren't
> > picking up the addition of _GNU_SOURCE (including stopping installing
> > the other tests in the same directory), and there's a bunch of tests
> > which #define _GNU_SOURCE in their code and now trigger build warnings.
> > I'm looking at fixes and mitigations now.

> Would it be better to revert this for now and get this for now? I wouldn't
> want you to extra busy work to workaround this.

I have patches that both fix the build breakage (it turns out on closer
inspection that it's just ALSA that's impacted I think, the extensive
extra warning splat and some other stuff that's just updates in
mainline) was making the impact look wider) and deescalate it to a
#warning which I'm currently running through CI.

> > The build failures are taking out the ALSA tests entirely which has
> > caused my personal CI to explode badly :/

> This has been in next for a while and I didn't see any failures. These
> kind of changes are tricky. On second thought, I probably should have
> delayed picking this up.

Yeah, I'd been on holiday for literally a month (got home Monday night)
- once I saw the issues I realised I had registered them going past at
some point while I was away but hadn't really been paying attention to
what was going on in -next.  I'm kind of surprised nobody else picked up
on the warnings though.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: -D_GNU_SOURCE kselftest breakage in mainline

[Shuah Khan]

On 5/16/24 09:09, Mark Brown wrote:
> On Thu, May 16, 2024 at 08:53:52AM -0600, Shuah Khan wrote:
>> On 5/16/24 08:02, Mark Brown wrote:
> 
>>> I'm seeing quite a lot of breakage in mainline as a result of
>>> daef47b89efd0b7 ("selftests: Compile kselftest headers with
>>> -D_GNU_SOURCE") and daef47b89efd0 ("selftests: Compile kselftest headers
>>> with -D_GNU_SOURCE") - thus far I've found that the use of
>>> static_assert() is triggering build breaks where testsuites aren't
>>> picking up the addition of _GNU_SOURCE (including stopping installing
>>> the other tests in the same directory), and there's a bunch of tests
>>> which #define _GNU_SOURCE in their code and now trigger build warnings.
>>> I'm looking at fixes and mitigations now.
> 
>> Would it be better to revert this for now and get this for now? I wouldn't
>> want you to extra busy work to workaround this.
> 


I determined the best option now is reverting the 3 patches.

commit daef47b89efd0b745e8478d69a3ad724bd8b4dc6
"selftests: Compile kselftest headers with -D_GNU_SOURCE"

"selftests/sgx: Include KHDR_INCLUDES in Makefile"
commit 2c3b8f8f37c6c0c926d584cf4158db95e62b960c

"selftests/cgroup: Drop define _GNU_SOURCE"
commit c1457d9aad5ee2feafcf85aa9a58ab50500159d2

I have this in linux-kselftest next now and will send pull
request to Linus this weekend.

Tejun, I made sure cgroup compiles after reverting the framework
patch and the cgroup test patch.

Mark,

Thank you Mark for finding the problem and sending in mitigation
fixes. I will hold off on your alsa patch:

kselftest/alsa: Ensure _GNU_SOURCE is defined
https://patchwork.kernel.org/project/linux-kselftest/patch/20240516-kselftest-fix-gnu-source-v1-1-e482ca6bfff7@kernel.org/

There is no need to for the following patch

kselftest: Desecalate reporting of missing _GNU_SOURCE
https://patchwork.kernel.org/project/linux-kselftest/patch/20240516-kselftest-mitigate-gnu-source-v1-1-a0e814ff2874@kernel.org/

thanks,
-- Shuah