dyung wrote:

> > No real updates here, but our internal builder did catch up to this commit 
> > and we are seeing the same (and a lot more) failures when this commit is 
> > merged into our downstream codebase. I was kind of hoping that it would 
> > pass so that it might indicate that the problem might be a configuration 
> > issue with our upstream build bot, but the fact that it fails in a similar 
> > manner likely points to some kind of bug in the version of MSVC that we are 
> > using.
> 
> It really would be helpful to try a different version of MSVC.
> 
> As I pointed out in the original reply to the build bot failure, this bot 
> seems to be using a version of MSVC that isn't even a listed release. The 
> last patch release of 16.9 is `192829923` according to Wikipedia, but this 
> builder reports `192829924`. To me, this seems to indicate a build that may 
> not have gone through full Q&A and such.

The build was released as far as I'm aware (we don't install pre-release 
versions on our production machines). That being said, I'm guessing it is 
probably a bug with the specific version of MSVC that we are using. I have 
started the conversation internally about upgrading the version of MSVC we use 
to build, but that will take time.

> Would it help for me to post a PR that adds some validation of the string 
> table? I can easily do that, and maybe that would let you isolate whether 
> this is definitely an MSVC version issue vs. something with the code.

Sure, I'm happy to try anything that you think might help and this is one of my 
higher priority issues at the moment.

> > @chandlerc could we consider reverting the change until we can figure out 
> > what the problem is? I will continue to take a look into it, but I don't 
> > really have any ideas what might be the cause at the moment, so have no ETA 
> > on a work-around, and in the meantime, this will continue to cause test 
> > failures for us internally.
> 
> I'm happy to revert, and I'll start working on that. But I think there needs 
> to be a concrete ETA on being able to make progress. Especially given the 
> fact that no other builders seem to hit this and it may be specific to an 
> MSVC version that even other windows devs don't realistically have access 
> to...
> 
> Can you give some fairly concrete timeline? 1 week? maybe 2? Otherwise, there 
> is no actual path to resolving this.

I will look into it the rest of this week and into next week. I don't really 
know the problem here, so I cannot promise a resolution date, but I will try my 
best to get this resolved, it is making our internal automation fail, so I want 
to unblock that!

https://github.com/llvm/llvm-project/pull/118734
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to