Testing a patch

Testing a patch

[Benjamin Priour]

Hi,

Regstrapping finally done for PR109439 - Spurious
-Wanalyzer-use-of-uninitialized-value tagging along
-Wanalyzer-out-of-bounds.
Now only a OOB warning is reported when necessary instead of OOB + Use
of uninitialized value.

Some tests in analyzer (out-of-bounds-*, realloc-5, pr101962) were
checking for the now removed use-of-uninitialized-value warning, and
therefore I fixed that.

But now I'm confused since the documentation reads that to perform
regtesting, one should use make -k check,
and that's what I always use too, but because I fixed the above test
files, contrib/compare_tests obviously complains about them having
disappeared.
Does it mean regtesting failed ? Can I submit the patch in its current
state or should I do something else before doing so ?
Guess I would get feedback anyway if something's wrong.

I figured I would send it here rather than to gcc-patches, as it's
more general than a discussion over a single patch.

Thanks,
Benjamin

Re: Testing a patch

[David Malcolm]

On Mon, 2023-05-29 at 15:09 +0200, Benjamin Priour wrote:
> Hi,
> 
> Regstrapping finally done for PR109439 - Spurious
> -Wanalyzer-use-of-uninitialized-value tagging along
> -Wanalyzer-out-of-bounds.
> Now only a OOB warning is reported when necessary instead of OOB +
> Use
> of uninitialized value.
> 
> Some tests in analyzer (out-of-bounds-*, realloc-5, pr101962) were
> checking for the now removed use-of-uninitialized-value warning, and
> therefore I fixed that.
> 
> But now I'm confused since the documentation reads that to perform
> regtesting, one should use make -k check,
> and that's what I always use too, but because I fixed the above test
> files, contrib/compare_tests obviously complains about them having
> disappeared.
> Does it mean regtesting failed ?

If part of the purpose of a patch is the removal of certain tests, then
if compare_tests is "complaining" about them disappearing, then
arguably compare_tests is in fact verifying that the behavior of the
patch matches the intended behavior.  In this case, regression testing
has effectively passed.


>  Can I submit the patch in its current
> state or should I do something else before doing so ?

Yes, please submit the patch in its current state, to gcc-patches, and
CC me on it.

Thanks
Dave

> Guess I would get feedback anyway if something's wrong.
> 
> I figured I would send it here rather than to gcc-patches, as it's
> more general than a discussion over a single patch.
> 
> Thanks,
> Benjamin
>