Repository navigation
replace TypingEnv::post_analysis() with mir::Body::typing_env() - #164077
adwinwhite wants to merge 1 commit into
Conversation
rustbot
added
S-waiting-on-author
This comment has been minimized.
This comment has been minimized.
2d08b00 to
0b642e2
Compare
This comment has been minimized.
This comment has been minimized.
0b642e2 to
53c2b35
Compare
| match self.phase { | ||
| MirPhase::Built | MirPhase::Analysis(_) => TypingEnv::new( | ||
| match self.typing_mode { | ||
| BodyTypingMode::Borrowck | BodyTypingMode::PostBorrowck => TypingEnv::new( |
There was a problem hiding this comment.
when do we encounter this?
There was a problem hiding this comment.
The initial mode is BodyTypingMode::Borrowck. After mir_borrowck, we set it to PostBorrowck.
There was a problem hiding this comment.
promote_consts pass would use BodyTypingMode::Borrowck since it's run before borrowck.
| /// us to see the difference and forego optimization on the inlined promoted items. | ||
| pub phase: MirPhase, | ||
|
|
||
| pub typing_mode: BodyTypingMode, |
There was a problem hiding this comment.
comment please :>
| StableHash, | ||
| TypeFoldable, | ||
| TypeVisitable | ||
| )] |
There was a problem hiding this comment.
should also have a comment, or maybe just change the field to go "see [BodyTypingMode] for more info"
also, mind splitting the derive in two so that it doesn't get spread out like this
5e23058 to
23a44a9
Compare
rustbot
added
the
S-waiting-on-review
|
Some changes occurred to MIR optimizations cc @rust-lang/wg-mir-opt |
rustbot
removed
the
S-waiting-on-author
This comment has been minimized.
This comment has been minimized.
| #[derive(Copy, Clone, PartialEq, Debug)] | ||
| #[derive(TyEncodable, TyDecodable, StableHash, TypeFoldable, TypeVisitable)] | ||
| pub enum BodyTypingMode { | ||
| Borrowck, |
There was a problem hiding this comment.
| // This branch happens for drop glue and fn ptr shims. | ||
| // FIXME: why do we do any of this analysis on drop glue etc? | ||
| // This should ideally all be skipped. | ||
| TypingEnv::post_analysis(tcx, self.source.def_id()) |
There was a problem hiding this comment.
should this be the case or could we change drop glue and fn ptr shims to just start out in TypingMode::PostAnalysis?
There was a problem hiding this comment.
We already use PostAnalysis mode for them. I guess this branch is actually useless.
|
|
||
| debug!("runtime_mir_cleanup({:?})", did); | ||
| run_runtime_cleanup_passes(tcx, body); | ||
| assert!(body.typing_mode == BodyTypingMode::PostAnalysis, "{:?}", body.typing_mode); |
There was a problem hiding this comment.
| // We try to guard against future MIR building procedures that don't go through the usual | ||
| // lowering. The final validation is expected to run in `TypingMode::PostAnalysis`. | ||
| if new_phase == MirPhase::Runtime(RuntimePhase::Optimized) { | ||
| assert_eq!(body.typing_mode, BodyTypingMode::PostAnalysis,); |
There was a problem hiding this comment.
| assert_eq!(body.typing_mode, BodyTypingMode::PostAnalysis,); | |
| assert_eq!(body.typing_mode, BodyTypingMode::PostAnalysis); |
|
@rustbot author |
rustbot
added
S-waiting-on-author
|
Reminder, once the PR becomes ready for a review, use |
23a44a9 to
a5cc7e2
Compare
|
This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed. Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers. |
This comment has been minimized.
This comment has been minimized.
a5cc7e2 to
4bbe3be
Compare
Similar to #163745
We should discourage constructing a typing env by hand when it can be derived from context.
This simplifies reasoning and reduces the mistake of not resetting rigidness markers across typing envs.
This PR adds a field to
mir::Bodyto track the typing mode.r? lcnr