[yashaggarwal85d]

KnownPanicsLint drops values of locals that are assigned more than once at the end of each block.
With overflow checks on, x >>= n puts an assert before the shift, so the shift ends up in the next
block and the lint loses x right there. This keeps visiting into the assert/goto target when it has
only one predecessor, and clears those locals once that chain ends.

Same applies to /= and %= (div-by-zero assert), which were missed even with -O.

Fixes #164191

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[rustbot]

r? @JonathanBrouwer

rustbot has assigned @JonathanBrouwer.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler, mir, mir-opt
• compiler, mir, mir-opt expanded to 77 candidates
• Random selection from 19 candidates

[oli-obk]

Reopening after a conversation with the author

[JonathanBrouwer]

@rustbot reroll
Not familiar enough with this code to feel comfortable reviewing