[llvm] [DebugInfo] Change return type of DWARFDebugAbbrev::parse (PR #67191)

[Alex Langford via llvm-commits]

bulbazord wrote:

> In principle, seems reasonable to me, though I worry slightly about the impact on existing users who won't expect this breaking-if-they-have-appropriate-options-selected change. How confident are you that you've caught all in-tree uses?

I would say I lean more towards confident than not. I've built clang, lld, lldb, bolt, and compiler-rt (all on an arm64-apple-macOS machine). The call-sites that I updated were found through a combination of commenting out the method in the header to introduce a compilation failure, IDE tooling, and tests. If there are places I did not catch, it is because of conditional compilation checks that my build did not meet. As always, if and when I land this I will pay attention to the CI machines.

I assume downstream users will adjust accordingly.

> Also, does this change allow us to check the lower-level error in a test somewhere, e.g. in a unit test?

Added a unit test similar to the other ones I introduced.

https://github.com/llvm/llvm-project/pull/67191