[PATCH v2 2/6] ld: Pass $NOPIE_CFLAGS and $NOPIE_LDFLAGS to test pr19719

[Jan Beulich]

On 24.03.2025 18:03, Jens Remus wrote:
> 
> 
> On 24.03.2025 17:25, Jan Beulich wrote:
>> On 24.03.2025 17:19, Jens Remus wrote:
>>> On 24.03.2025 17:07, Jan Beulich wrote:
>>>> On 21.03.2025 17:58, Jens Remus wrote:
>>>>> Linker test "pr19719 fun defined" (non PIE) fails on s390x on Fedora 41
>>>>> but not on Ubuntu 24.10.  The reason is that GCC on Ubuntu is configured
>>>>> with --enable-default-pie, so that it defaults to compile with -fPIE
>>>>> and link with -pie, which hides the test fail.
>>>>
>>>> I'm confused: Generally I'd expect all tests to pass. Yet "hides the test
>>>> fail" suggests the test is actually "expected" to fail on, however without
>>>> it being an xfail. What am I missing?
>>>
>>> This test currently fails on s390x on distributions that configure GCC
>>> without "--enable-default-pie", but does not on distributions that do.
>>> As far as I can tell the test expects to be linked as PDE, but gets
>>> linked as PIE if GCC is configured with with "--enable-default-pie".
>>> For PDE this is a known test fail on s390x and I am working on resolving
>>> all of the ld testsuite fails on s390x. Before that I would like to
>>> make the test run deterministic.
>>
>> I'd like to leave approval to an s390x maintainer then.
> 
> @Andras: Could you please approve?
> 
> @Jan: My change is not s390x-specific.

Sure. Yet still I don't feel comfortable approving it without having a way
to verify the assertion of the testcase's status on s390x.

>  Don't you in general agree that
> test pr19719 expects to be linked as PDE and that test pr19719pie
> expects to be linked as PIE (as the name clearly states)?

The latter name is clear; the former is ambiguous. It may have been deemed
unambiguous at a time where PIE was still something new, unusual.

>  Similar to my
> other linker test changes in this series I am just nullifying whether
> GCC defaults to compile with -fPIE and link with -pie.  Otherwise tests
> pr19719 and pr19719pie would be identical when GCC defaults to PIE. That
> does not make any sense to me.
> 
> Note that this does not change that test pr19719 fails on distributions
> that don't configure GCC to default to PIE.  But that is a separate
> issue that I am working on separately.

All fine, yet as per above - you will need to allow me to not feel
comfortable to approve this or effectively any change.

Jan