Improve liveness analysis for generators

[tmiasko]

Liveness analysis for generators assumes that execution always continues
normally after a yield point, not accounting for the fact that generator
could be dropped before completion.

If generators captures any variables by reference, those variables could
be used within a generator, or when the generator completes, but also
after each yield point in the case the generator is dropped.

Account for the case when generator is dropped after yielding, but
before running to the completion. This effectively considers all
variables captured by reference to be used after a yield point.

Fixes #84292.

[rust-highfive]

r? @estebank

(rust-highfive has picked a reviewer for you, use r? to override)

[tmandry]

r? @tmandry

cc @eholk fyi

[eholk]

This looks good to me. To make sure I understand it, it looks like the main thing this does is adds an edge from every yield expression to the exit node, which models the fact that execution may not resume after a yield. Is this correct?

[tmiasko]

it looks like the main thing this does is adds an edge from every yield expression to the exit node, which models the fact that execution may not resume after a yield. Is this correct?

Yes, that's correct. Variables captured by reference are already marked as live in the exit node at the beginning of Liveness::compute. To fix the issue we just need to model the control flow from yield to the exit.

[bors]

☔ The latest upstream changes (presumably #85556) made this pull request unmergeable. Please resolve the merge conflicts.

[tmiasko]

Rebased to resolve conflicts with #85556.

[tmandry]

Looks good, thanks!

@bors r+

[bors]

📌 Commit 9f6f862 has been approved by tmandry

[bors]

⌛ Testing commit 9f6f862 with merge 1a9ac38...

[bors]

☀️ Test successful - checks-actions
Approved by: tmandry
Pushing 1a9ac38 to master...

[tmiasko]

Thanks for reviewing @eholk, @tmandry.