On 11/20/25 22:51, Jonathan Wakely wrote:
On Thu, 20 Nov 2025, 20:36 François Dumont, <[email protected]> wrote:
On 11/19/25 14:51, Jonathan Wakely wrote:
> On Tue, 18 Nov 2025 at 21:38, François
Dumont<[email protected]> wrote:
>> Ok, here is the new patch, only fixing thestd::erase_if
behavior for
>> the _GLIBCXX_DEBUG.
> Is that right? I thought this change is to benefit direct uses of
> __gnu_debug::vector notstd::vector with _GLIBCXX_DEBUG? The
first new
> test already passes when usingstd::vector with -D_GLIBCXX_DEBUG. In
> fact, it FAILs with this patch! So this change is not helping
> _GLIBCXX_DEBUG, it's making it worse.
>
> I see the problem there, you need to check __glibcxx_erase_if in
> <debug/vector>, not __cpp_lib_erase_if. The latter hasn't been
defined
> yet because <debug/vector> gets included before
> __glibcxx_want_erase_if is defined in <vector>. You probably
won't see
> the FAILs for a default configuration of GCC, but if you configure
> with --disable-libstdcxx-pch then it will fail.
Caught again by pch, I wonder if those should not be disabled per
default when running tests.
I did this for the 2 new tests.
We don't want the no_pch option in these tests, they're not special,
and they're not testing macros which really do depend on pch.
These tests (and 99% of the tests in the testsuite) should work
correctly with pch or without pch.
Ok, I'll stick to that plan in the future.
Thanks for taking care.
