[libcxx-dev] Libcxx testing with top-of-tree clang?

[David Zarzycki via libcxx-dev]

> On Dec 1, 2019, at 11:59 PM, Marshall Clow <mclow.lists at gmail.com> wrote:
> 
> On Fri, Nov 29, 2019 at 11:51 PM David Zarzycki <dave at znu.io <mailto:dave at znu.io>> wrote:
>> On Nov 29, 2019, at 11:30 PM, Marshall Clow <mclow.lists at gmail.com <mailto:mclow.lists at gmail.com>> wrote:
>> 
>> On Fri, Nov 29, 2019 at 3:50 AM David Zarzycki via libcxx-dev <libcxx-dev at lists.llvm.org <mailto:libcxx-dev at lists.llvm.org>> wrote:
>> Do any of the bots test libcxx with top-of-tree clang? More than once I’ve suspected the answer is “no”. Most recently, the test suite has been broken for a few days now. I’ve worked around the problem by adding -Wno-deprecated-copy to the test suite config but somebody needs to decide if that’s the best long term solution.
>> 
>> My understanding is that (several of) the libcxx test bots get updated to the latest clang on a regular basis, but not on every run.
>> So they may lag behind ToT by a week or so at times.
> 
> Okay.
> 
> Separably, why is important for the test suite to fail when new warnings are introduced? Is there a way to relax the test suite?
> 
> I believe that it is the responsibility of the people implementing the new warnings to ensure that all of LLVM (and that includes clang, libc++, etc) builds cleanly with the new warning(s) enabled - and that this should happen *before* the warning is enabled.
> 
> This is what I have done when I have made changes to clang.

I think my question was misunderstood, so please let me rephrase: why is important that the test suite pass with all warnings (both present and future) turned into errors?

At the very least, can we please add a CMake option to disable unqualified -Werror? (i.e. -Werror without the equal sign.)

Dave

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/libcxx-dev/attachments/20191202/65383946/attachment.html>