Skip to content

Conversation

@fuzyll
Copy link
Contributor

@fuzyll fuzyll commented Oct 16, 2025

After some DMs with @negasora, this should hopefully address #7480.

The problem is that we're not applying relocations. So, the data we're giving to gimli is incorrect. We need to apply those relocations.

Unfortunately, since we're in a raw view, that's not trivial to do. An earlier version of this code tried to apply the relocations ourselves, but it turns out we can't even get the correct relocation ranges out of the BV at this point.

We had the idea of taking on object as a dependency to tell us where the relocations are and help us apply them manually before we hand stuff off to gimli. This feels bad, though, so I tried to walk Claude Code through the problem in the hopes that it would help us solve it without needing object.

After about 30 minutes, it decided to give up and take the dependency itself despite my having told it not to. So, I've thrown in the towel, and here's a PR with a cleaned-up version of the solution.

Feel free to reject and do something else if you can, but this code does work for the crashing sample. I have it going through CI right now, so we'll see if it survives other platforms and unit tests, too.

@negasora
Copy link
Member

Closing since b6740d0 got merged

@negasora negasora closed this Oct 17, 2025
@fuzyll fuzyll deleted the test_fix_dwarf_relocation_bug branch October 18, 2025 23:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants