Without the patch, all version 5 compile units in a DWP file read location tables from the beginning of a .debug_loclists.dwo section. The patch fixes that by adjusting the reading offset the same way as for pre-v5 units. The section identifier to find the contribution entry corresponds to the version of the unit.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Again, looks good, although I'm not up-to-speed on the .debug_loc/.debug_loclists format, so this should get a second pair of eyes on it.
llvm/test/DebugInfo/X86/dwp-v2-loc.s | ||
---|---|---|
2–4 | Was this not already tested? |
llvm/test/DebugInfo/X86/dwp-v2-loc.s | ||
---|---|---|
2–4 | I was thinking the same thing. The v2 version of this patch (https://reviews.llvm.org/D53155) did include a test case, but that one was based on running llc+dwp. I think that this kind of a test is preferable for testing dumping. I was thinking about deleting the other test, but then then I thought, maybe it is still valueable as it also tests llvm-dwp? |
llvm/test/DebugInfo/X86/dwp-v2-loc.s | ||
---|---|---|
2–4 | Ah, yep - I'll blame my past lazy self. I've not always been good about doing the Right Thing and adding the llvm-dwarfdump functionality separately (& testing it with object files or assembly tests) then adding the production functionality (& testing that with IR+llvm-dwarfdump). Yep, both testing makes sense. |
Was this not already tested?