Misc cleanups from reading some borrowck code #150141
Conversation
rustbot
added
S-waiting-on-review
| let identity_args = GenericArgs::identity_for_item(tcx, typeck_root_def_id); | ||
| if self.mir_def.to_def_id() == typeck_root_def_id | ||
| // Do not ICE when checking default_field_values consts with lifetimes (#135649) | ||
| && DefKind::Field != tcx.def_kind(tcx.parent(typeck_root_def_id)) |
There was a problem hiding this comment.
| let arg_mapping = iter::zip(identity_args.regions(), fr_args.regions().map(|r| r.as_var())); | ||
| // This relies on typeck roots being generics_of parents with their | ||
| // parameters at the start of nested bodies' generics. | ||
| assert!(renumbered_args.len() >= identity_args.len()); |
There was a problem hiding this comment.
I don't know how valuable this assert is in practice? I also would expect us to be relying on this invariant in more places in the compiler than just here 🤔 feels like more generally we expect all typeck parents to be generics_of parents and if that doesn't hold we probably want to yell somewhere other than here 🤔
|
|
||
| mod opaque; | ||
|
|
||
| fn anon_const_type_of<'tcx>(icx: &ItemCtxt<'tcx>, def_id: LocalDefId) -> Ty<'tcx> { |
There was a problem hiding this comment.
ive just moved this down because it's weird for type_of to be sandwiched between helper functions both on top and below :3 no actual changes here
This comment has been minimized.
This comment has been minimized.
BoxyUwU
force-pushed
the
borrowck_cleanup_1
branch
from
9ef21f0 to
9304256
Compare
| ); | ||
| DefiningTy::Const(self.mir_def.to_def_id(), args) | ||
| } else { | ||
| // FIXME this line creates a dependency between borrowck and typeck. |
There was a problem hiding this comment.
| // This line creates a dependency between borrowck and typeck. |
Want to remove this? I don't get this FIXME. THere's nothing to fix here 🤣
There was a problem hiding this comment.
I think this about query edges, could imagine encoding the type of inline consts in the MIR somehow so that u only depend on the mir query rather than typeck
4 participants
title
r? lcnr