@bors try @rust-timer queue

☀️ Try build successful (CI)
Build commit: d38d65d (d38d65d55029fcaeb492c32d33ea24894520d7de, parent: 0c68443b0a0469e4211acca7e7b06e14f256ada8)

Whoops, it turns out that my previous benchmark run had a bug that would cause it to simply not perform certain ChunkedBitSet bulk updates at all, which is blatantly incorrect. 😅

This implies that I should probably also add some relevant unit tests.

Let's try again with a version that actually performs real work.

@bors try @rust-timer queue

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

⌛ Trying commit 20b8c98 with merge cffefe3…

To cancel the try build, run the command @bors try cancel.

Workflow: https://github.com/rust-lang/rust/actions/runs/22937404624

Now there’s some kind of bug that causes the modified compiler to hang while building proc_macro.

@bors try cancel

Try build cancelled. Cancelled workflows:

• https://github.com/rust-lang/rust/actions/runs/22937404624

Closing this for now, as it needs better testing/debugging.

Ah, I think the bug might have been in unconditionally returning changed = true from the new fast path, when in fact it can sometimes result in no modification.

This would then break code that relies on the return value when iterating to fixpoint, because it would incorrectly believe that fixpoint is never reached, and loop forever.