rupprecht wrote:

> Can you find the template that causes this issue and reduce it down to a test 
> similar to the test for this PR?

The problematic type is `Future<int>` from 
https://github.com/google/tensorstore/blob/master/tensorstore/util/future.h, 
but the data formatter for it is internal, and might be necessary to reproduce. 
(The test case also exercises `Future<void>`, and that was fine). I'll be out 
next week though, so unless I get lucky in attempting to reduce this in the 
next day, it will be a while until I have something. @cmtice might be able to 
help if you want something sooner.

> So the PR didn't introduce a regression — it made an existing unsupported 
> case visible in dev builds instead of silently wrong.

Well, I slightly disagree here. Crashing in dev builds is still a form of 
regression, even if prod builds are unaffected.

To be clear though, I'm totally fine w/ how this patch landed -- it's great 
that you added the assert to catch this issue. I assume we believed that all 
the cases were handled, and an assert is a great way to either confirm that or 
to discover when that is actually not the case.

But... once there's an assertion failure, I think we _do_ need to call that a 
regression of some sort, and resolve it either by handling that case so we 
don't get to the assertion, or removing the assertion if we can't figure that 
out yet. It is probably fine to re-land this as-is but w/o the assert (assuming 
you resolve the windows build bot issue).

https://github.com/llvm/llvm-project/pull/187598
_______________________________________________
lldb-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits

Reply via email to