[PATCH] Fix libstdc++ testsuite to handle VxWorks gthreads implementation

Jonathan Wakely jwakely@redhat.com
Tue Aug 25 07:48:01 GMT 2020

On 24/08/20 23:19 -0300, Alexandre Oliva wrote:
>On Aug 24, 2020, Jonathan Wakely <jwakely@redhat.com> wrote:
>>>> OK for trunk, thanks.
>>> Given the approval and the lack of significant changes, I'll put this in
>>> unless there are objections in the next 48 hours.  Thanks for the review!
>> Thanks. There's a new FAIL due to a bad merge.
>Erhm...  Weird, I don't recall any manual adjustments to
>packaged_task/cons/alloc.cc.  Indeed, the changes look exactly the same
>that Corentin had proposed before.  Maybe the constraints were already
>there, but hte unintended effects were not noticeable before the change
>to the default?

You're right, it's not a bad merge (just a bad change I should have
noticed in the review :-)

My change to prevent it running for C++17 was done May 2019 with
9a0af7e3fb425ae2c0e044d044feb81ef493ce2c so that was already there
before Corentin's original patch.

>Anyway, thanks for fixing the failure.  That said, I confess I don't get
>how this test will (or should) ever be run in C++14 mode, being given an
>explicit -std=gnu++11.  Is this really the right way to go about it?

The medium term plan is to run libstdc++ tests with several different
-std options, as is done for g++ tests. At that point we can drop most
(but not all) explicit -std options in the tests. But until then, this
test won't run at all if we don't put an explicit -std there, because
it gets skipped with the default -std=gnu++17.

Testing with -std=gnu++11 is better than not testing at all, and the
code being tested is identical for C++11 and C++14 so it's not a
problem that it isn't tested for C++14.

Now and then I manually remove all -std options from the tests and run
them with multiple -std permutations, so it does get tested with C++14
mode, just not automatically.

More information about the Libstdc++ mailing list