Turns out
- `owning_ref` is unsound due to `Box` aliasing stuff
- `rustc` doesn't need 99% of the `owning_ref` API
- `rustc` can use a far simpler abstraction that is `OwnedSlice`
Tweak debug outputs to make debugging new solver easier
1. Move the fields that are "most important" (I know this is subjective) to the beginning of the structs.
For goals, I typically care more about the predicate than the param-env (which is significantly longer in debug output).
For canonicalized things, I typically care more about what is *being* canonicalized.
For a canonical response, I typically care about the response -- or at least, it's typically useful to put it first since it's short and affects the whether the solver recurses or not...
2. Add some more debug and instrument calls to functions to add more structure to tracing lines.
r? `@oli-obk` or `@BoxyUwU` (since I think `@lcnr` is on holiday)
submodule detection for proper fix on #96188
This commit resolves an internal FIXME note within the bootstrap by implementing submodule detection. This is accomplished through an iterative process over the `.gitmodules` file.
r? `@albertlarsan68`
Add T-bootstrap label to tools
cc `@rust-lang/bootstrap:`
### Unresolved questions:
- [ ] Do we want to keep the A-testsuite on those?
- [ ] Do we want all of those tools to be ours?
This commit resolves an internal FIXME note within the bootstrap by implementing submodule detection.
This is accomplished through an iterative process over the `.gitmodules` file.
Signed-off-by: ozkanonur <work@onurozkan.dev>
Avoid a few locks
We can use atomics or datastructures tuned for specific access patterns instead of locks. This may be an improvement for parallel rustc, but it's mostly a cleanup making various datastructures only usable in the way they are used right now (append data, never mutate), instead of having a general purpose lock.
Remove the use of `-use-gnu-stack` when BOLTing LLVM
This flag was (counterintuitively) removing the `GNU_STACK` ELF attribute, which caused the optimized `libLLVM.so` file to be flagged as having an executable stack on SELinux.
Removing the flag might cause issues with `strip`. I'm not aware that we're stripping `libLLVM.so` though. Does it happen anywhere?
Fixes: https://github.com/rust-lang/rust/issues/105783
Disable path trimming during graphviz output
Fixes#109943
We can end up pretty-printing a path for the graphviz file without emitting a diagnostic, so disable path trimming at this point.
Allow `transmute`s to produce `OperandValue`s instead of needing `alloca`s
LLVM can usually optimize these away, but especially for things like transmutes of newtypes it's silly to generate the `alloc`+`store`+`load` at all when it's actually a nop at LLVM level.
LLVM can usually optimize these away, but especially for things like transmutes of newtypes it's silly to generate the `alloc`+`store`+`load` at all when it's actually a nop at LLVM level.
Do not suppress temporary_cstring_as_ptr in macros.
There isn't really a reason to skip the lint when part of the expression comes from an expansion.
Fixes https://github.com/rust-lang/rust/issues/94694
This flag (counterintuitively) was removing the `GNU_STACK` ELF attribute, which caused the optimized `libLLVM.so` file to be flagged as having an executable stack on SELinux.
Rollup of 9 pull requests
Successful merges:
- #109723 (Pull some tuple variant fields out into their own struct)
- #109838 (Fix `non_exhaustive_omitted_patterns` lint span)
- #109901 (Enforce VarDebugInfo::Place in MIR validation.)
- #109913 (Doc-comment `IndexVec::from_elem` and use it in a few more places)
- #109914 (Emit feature error for parenthesized generics in associated type bounds)
- #109919 (rustdoc: escape GAT args in more cases)
- #109937 (Don't collect return-position impl traits for documentation)
- #109938 (Move a const-prop-lint specific hack from mir interpret to const-prop-lint and make it fallible)
- #109940 (Add regression test for #93911)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Move a const-prop-lint specific hack from mir interpret to const-prop-lint and make it fallible
fixes#109743
This hack didn't need to live in the mir interpreter. For const-prop-lint it is entirely correct to avoid doing any const prop if normalization fails at this stage. Most likely we couldn't const propagate anything anyway, and if revealing was needed (so opaque types were involved), we wouldn't want to be too smart and leak the hidden type anyway.
Don't collect return-position impl traits for documentation
#104889 modified the rustdoc ast collection step to use a HIR visitor, which more thoroughly walks the HIR tree. that means that we're going to encounter inner items (incl return-position impl traits and async fn opaque futures) that are not possible to document.
FIxes (but does not close due to being a beta regression) #109931
r? `@GuillaumeGomez`
rustdoc: escape GAT args in more cases
Fixes#109488.
Previously we printed the *un*escaped form of GAT arguments not only when `f.alternate()` was true but *also* when we failed to compute the URL of the trait associated with the type projection, i.e. when `href(…)` returned an `Err(_)`.
In this PR the argument printing logic is entirely separate from the link resolution code above as it should be.
Further, we now only try to compute the URL if the HTML format was requested with `!f.alternate()`. Before, we would sometimes compute the `href` only to throw it away later.
