[PATCH] D41416: [modules] [pch] Do not deserialize all lazy template specializations when looking for one.

Thu Jul 13 01:05:35 PDT 2023

ChuanqiXu added a comment.

In D41416#4496293 <https://reviews.llvm.org/D41416#4496293>, @v.g.vassilev wrote:

> In D41416#4492367 <https://reviews.llvm.org/D41416#4492367>, @v.g.vassilev wrote:
>
>> Address most of the comments. I will need some help with the `Modules/pr60085.cppm` failure. I suspect we pass to CodeGen some instantiation by iterator index and that does not work as the instantiations are lazily triggered upon use now.
>
> In fact it looks like it fails exactly in the same way as described in the original report https://github.com/llvm/llvm-project/issues/60085. The commit message in https://github.com/llvm/llvm-project/commit/78e48977a6e67 hints at the fact that the issue was gone surprisingly. I suspect that test case stopped reproducing the underlying issue by chance. That probably means that this patch is not breaking anything but exposing an underlying problem...

Got it. I understand it is frustrating to fix an additional bug during development. Also it is generally not good to revert a valid test case.

Are you in a hurry to land this recently? (e.g., land this for clang17) If not, I suggest to wait for me to fix the underlying issue after clang17 gets released. And if I take too long to fix it (e.g, 9.30), given the speciality of the patch, I think we can revert that test case and land this one that time. How do you feel about the idea?

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D41416/new/

https://reviews.llvm.org/D41416