[DWARFLinker] Use different addresses to distinguish invalid DW_AT_LLVM_stmt_sequence offset

[DataCorrupted]

It'd be helpful (especially when llvm-dwarfdump ... | grep <invalid_address>) to separate two different invalid reasons for debugging.

[ellishg]

Could we add a test?

[DataCorrupted]

The whole debugging process would be much easier if I have a min repro when either is generated. So no at the moment. By the time I can add a test, I would be close to resolving this :\

[JDevlieghere]

@DataCorrupted Please don't remove reviewers when they get added automatically. It defeats the purpose of the person creating those rules.

Regarding the test, I understand you're making the change to debug an issue that you don't have a reproducer for, but the the case where the origin offset is missing seems like a situation for which you could hand-craft a test case?

[DataCorrupted]

@JDevlieghere Thanks for the advice.

Yes I could add a case where the original offset is missing, but that's not the bug I'm working on. The bug I have here is that the object file contains valid offsets, but that offset went missing after linking. I'm still chasing when/where that offset went missing. I'll add a test when I can create a repro for that, and include you as the reviewer.