[Clang] Fix an integer overflow issue in computing CTAD's parameter depth

[zyn0217]

There were some cases where we computed incorrect template parameter depths for synthesized CTAD, invalid as they might be, we still shouldn't crash anyway.

Technically the only scenario in which the inner function template's depth is 0 is when it lives within an explicit template specialization, where the template parameter list is empty.

Fixes #128691

[hokein]

Thanks for the quick fix!

[hokein]

IIUC, for fully explicit specializations where the template parameter list is empty <>, they don't contribute to the depth. This makes me think we should use Depth directly if FTD->getDeclContext() is an explicit specialization?

Would it be possible the Depth is not 0 in that case? an explicit template specialization is inside a template class?

[zyn0217]

Hmm interesting question. Afaict, the transform ConvertConstructorToDeductionGuideTransform only handles class templates nested no more than one level. So basically:

• For one-level nested cases (e.g., https://godbolt.org/z/x94dzMaeo), the original template parameter depth is 1, so the new depth would be definitely 0, which seems correct.
• For multi-level nested cases, we synthesize a CTAD guide with all templates except the last level being instantiated, effectively reducing it to the one-level scenario.

So I think that, in the end the depth should always be 0. But I'm happy to use Depth instead of a literal 0 if that seems more appropriate.

[zyn0217]

Oh I realized we sometimes still want to use Depth - 1 even if it is an explicit specialization (because it could be nested as in the first case, and the transformed depth should be 0) - so I think this is good enough?

[hokein]

So I think that, in the end the depth should always be 0. But I'm happy to use Depth instead of a literal 0 if that seems more appropriate.

I agree, this seems like always the case, Depth for the new transformed template parameter is 0.

I think we need some comments for the special 0 case.

[hokein]

(We should have been doing this wrong prior to 20, but it didn't crash because it called Decl::getTemplateDepth(), which returns 0 for the transformed template parameter declaration)

I think you're right. The depth is already incorrect in clang19 etc. https://godbolt.org/z/d9xhxh3EM

 |-TemplateTypeParmDecl 0x175b5b88 <line:6:13, col:22> col:22 typename depth 32767 index 1 T

[hokein]

Need to add a release note as it fixes an existing issue.

[hokein]

So I think that, in the end the depth should always be 0. But I'm happy to use Depth instead of a literal 0 if that seems more appropriate.

I agree, this seems like always the case, Depth for the new transformed template parameter is 0.

I think we need some comments for the special 0 case.