[llvm] [DwarfGenerator] Calculate relative offset according to Dwarf Version (PR #84847)

[Will Hawkins via llvm-commits]

hawkinsw wrote:

> LGTM, not sure how this isn't already causing issues...!
> 

I had exactly the same thought. The reason there is no breakage, I think, is that it "mostly" works and if the unit tests are "round tripping" (ie, emit and read), it is "close enough" that it works. In particular, if you use llvm-dwarfdump on the emitted blobs, you get "mostly" good information. 

But, seeing this code made me have one of those classic programmer moments: "How did this ever work?"


> > Looks good. Can we add a unit test for this?
> 
> This is already code for generating test inputs and nothing else. What exactly would you test?

I also wondered that, but was/am willing to brainstorm and have a try at it! 

Thank you for the review!!



https://github.com/llvm/llvm-project/pull/84847