Previously we were also marking undefined symbols (i.e. imports)
as live.
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
Looks good. I'm deducing that when you removed the "hidden" from "used_undef_fn" and "unused_undef_fn", the test then failed - and then the code change makes it pass again. Makes sense.
We could also explicitly test everything - by adding used_undef_hidden_fn/used_undef_public_fn/unused_undef_hidden_fn/unused_undef_public_fn, and the same for the globals as well. Might be worth it, just to cover all the cases.
test/wasm/gc-imports.ll | ||
---|---|---|
3 ↗ | (On Diff #152186) | Could also remove -print-gc-sections here (or check/assert that its output matches what we expect). I think the checks were lost in all the debate about where and whether different section/function/global types would be reported... |
I've been quiet for a while as well ... I keep meaning to get back to my patch queue and try and get them reviewed and submitted. I have a couple of features completed or mostly-completed, I just need a few days away from my main job to get them back on track! Maybe it's time to get LLD ready for native-exception "sections" as well - that's a feature I'm really hoping will arrive soon in WebAssembly.