introduce canonical queries, use for normalization and dropck-outlives
This branch adds in the concept of a **canonicalized trait query** and uses it for three specific operations:
- `infcx.at(cause, param_env).normalize(type_foldable)`
- normalizes all associated types in `type_foldable`
- `tcx.normalize_erasing_regions(param_env, type_foldable)`
- like normalize, but erases regions first and in the result; this leads to better caching
- `infcx.at(cause, param_env).dropck_outlives(ty)`
- produces the set of types that must be live when a value of type `ty` is dropped
- used from dropck but also NLL outlives
This is a kind of "first step" towards a more Chalk-ified approach. It leads to a **big** speedup for NLL, which is basically dominated by the dropck-outlives computation. Here are some timing measurements for the `syn` crate (pre-branch measurements coming soon):
| Commit | NLL disabled | NLL enabled |
| ------- | --- | --- |
| Before my branch | 5.43s | 8.99s |
| After my branch | 5.36s | 7.25s |
(Note that NLL enabled still does *all the work* that NLL disabled does, so this is not really a way to compare the performance of NLL versus the AST-based borrow checker directly.) Since this affects all codepaths, I'd like to do a full perf run before we land anything.
Also, this is not the "final point" for canonicalization etc. I think canonicalization can be made substantially faster, for one thing. But it seems like a reasonable starting point for a branch that's gotten a bit larger than I would have liked.
**Commit convention:** First of all, this entire branch ought to be a "pure refactoring", I believe, not changing anything about external behavior. Second, I've tagged the most important commits with `[VIC]` (very important commit), so you can scan for those. =)
r? @eddyb
Before, the identifier `X` was also used when generating a pattern
to match against the dep-node. So `Foo(DefId)` would generate a match
pattern like:
match foo {
Foo(DefId) => ...
}
This does not scale to more general types like `&'tcx
Ty<'tcx>`. Therefore, we now require *exactly one* argument (the macro
was internally tupling anyway, and no actual nodes use more than one
argument), and then we can generate a fixed pattern like:
match foo {
Foo(arg) => ...
}
Huzzah. (Also, hygiene is nice.)
In general, we've been moving towards a semantics where you can have
contradictory where-clauses, and we try to honor them. There are
already existing run-pass tests where we take that philosophy as
well (e.g., `compile-fail/issue-36839.rs`). The current behavior of
`and`, where it strips the environment, breaks that code.
- `ParamEnv::empty()` -- does not reveal all, good for typeck
- `ParamEnv::reveal_all()` -- does, good for trans
- `param_env.with_reveal_all()` -- converts an existing parameter environment
Improvements:
- Use Clone not Copy for the "simple cases"
- Separate TypeFoldable and Lift for the "simple cases"
- Support generics type parameters
- Support named fields in enum variants
- etc
Remember state of top-level collapse toggle widget
This change allows the big top-right expand/collapse toggle to remember its setting across navigation or page reloads. Prior to this change, there was this annoyance:
- browse to some docs
- Click the minus button to collapse them
- browse to other docs (or reload the page)
- Everything is expanded again
The solution is based on storing a simple boolean flag in localStorage. I think it's a good improvement, but it does introduce the following potentially surprising behavior:
- browse to some docs
- click the minus button to collapse them
- click to expand a particular item (not the main top-right big one)
- reload the page, everything is collapsed
Paired with @debugsteven on this.
Fix hygene issue when deriving Debug
The code for several of the core traits doesn't use hygenic macros.
This isn't a problem, except for the Debug trait, which is the only
one that uses a variable, named "builder".
Variables can't share names with unit structs, so attempting to
[derive(Debug)] on any type while a unit struct with the name
"builder" was in scope would result in an error.
This commit just changes the name of the variable to
"__debug_trait_builder", because I couldn't figure out how to get a
list of all unit structs in-scope from within the derive expansion
function. If someone wants to have a unit struct with
the exact name "__debug_trait_builder", they'll just have to do it
without a [derive(Debug)].
I also checked the implementations of the other built-in derives to
ensure they didn't declare any variables.
in which some labels and notes are upgraded to structured suggestions
(Meanwhile, a couple of parse-fail tests are moved to UI tests so that
the reader can see the new output, and an existing UI test is given a
more evocative name.)
r? @estebank
Update L4Re target specification
Due to the dynamically generated linker arguments of the L4Re build system, it is not a good idea to hard-code them in Rust. This PR undoes this step. It also adds an empty implementation to retrieve the number of CPUs.
Update Feature Request instructions
As noted in #48393 the contribution instructions for submitting a Feature Request are a little hasty, suggesting that the user immediately create an issue in the RFC repository. For users that want to submit a feature request, let's instead point them directly to the README file for the RFC repository, which contains detailed instructions on how to submit a Feature Request.
Two phase borrows rewrite
This definitely needs a careful review. Both @pnkfelix and @nikomatsakis were involved with the design of this so they're natural choices here. I'm r?'ing @pnkfelix since they wrote the original two-phase borrow implementation. Also ping @KiChjang who expressed interest in working on this. I'm going to leave a few comments below pointing out some of the more dangerous changes I made (i.e. what I would like reviewers to pay special attention too.)
r? @pnkfelix
test: Forcibly remove MAKEFLAGS in compiletest
When executing run-make tests we run a risk of leaking the `MAKEFLAGS`
environment variable if `./x.py` itself was called from `make` (aka `make check
-j3` as the OSX bots do). We may then leak accidentally fds into the child
process and trick it into thinking it's got a jobserver!
Hopefully addresses [this] spurious failure
[this]: https://github.com/rust-lang/rust/pull/48295#issuecomment-372134717
When executing run-make tests we run a risk of leaking the `MAKEFLAGS`
environment variable if `./x.py` itself was called from `make` (aka `make check
-j3` as the OSX bots do). We may then leak accidentally fds into the child
process and trick it into thinking it's got a jobserver!
Hopefully addresses [this] spurious failure
[this]: https://github.com/rust-lang/rust/pull/48295#issuecomment-372134717
