rustc_metadata: track the simplified Self type for every trait impl.
For the `traits_impls_of` query, we index the impls by `fast_reject::SimplifiedType` (a "shallow type"), which allows some simple cases like `impl Trait<..> for Foo<..>` to be efficiently iterated over, by e.g. `for_each_relevant_impl`.
This PR encodes the `fast_reject::SimplifiedType` cross-crate to avoid needing to deserialize the `Self` type of every `impl` in order to simplify it - the simplification itself should be cheap, but the deserialization is less so.
We could go further from here and make loading the list of impls lazy, for a given simplified `Self` type, but that would have more complicated implications for performance, and this PR doesn't do anything in that regard.
r? @nikomatsakis cc @Mark-Simulacrum
Limit I/O vector count on Unix
Unix systems enforce limits on the vector count when performing vectored I/O via the readv and writev system calls and return EINVAL when these limits are exceeded. This changes the standard library to handle those limits as short reads and writes to avoid forcing its users to query these limits using platform specific mechanisms.
Fixes#68042
Show backtrace numbers in backtrace whenever more than one is involved
Previously, we only displayed 'frame' numbers in a macro backtrace when more
than two frames were involved. This commit should help make backtrace
more readable, since these kinds of messages can quickly get confusing.
Previously, we only displayed 'frame' numbers in a macro backtrace when more
than two frames were involved. This commit should help make backtrace
more readable, since these kinds of messages can quickly get confusing.
All #[cfg(unix)] platforms follow the POSIX standard and define _SC_IOV_MAX so
that we rely purely on POSIX semantics to determine the limits on I/O vector
count.
Keep the I/O vector count limit in a `SyncOnceCell` to avoid the overhead of
repeatedly calling `sysconf` as these limits are guaranteed to not change during
the lifetime of a process by POSIX.
Both Linux and MacOS enforce limits on the vector count when performing vectored
I/O via the readv and writev system calls and return EINVAL when these limits
are exceeded. This changes the standard library to handle those limits as short
reads and writes to avoid forcing its users to query these limits using
platform specific mechanisms.
Add regression test for #72787
Add regression test for Issue #72787Fixes#72787
~~Still waiting on running tests locally to bless the error output~~
r? @lcnr
polymorphization: various improvements
This PR includes a handful of polymorphisation-related changes:
- @Mark-Simulacrum's suggestions [from this comment](https://github.com/rust-lang/rust/pull/74633#issuecomment-668684433):
- Use a `FiniteBitSet<u32>` over a `FiniteBitSet<u64>` as most functions won't have 64 generic parameters.
- Don't encode polymorphisation results in metadata when every parameter is used (in this case, just invoking polymorphisation will probably be quicker).
- @lcnr's suggestion [from this comment](https://github.com/rust-lang/rust/pull/74717#discussion_r463690015).
- Add an debug assertion in `ensure_monomorphic_enough` to make sure that polymorphisation did what we expect.
r? @lcnr
Completes support for coverage in external crates
Follow-up to #74959 :
The prior PR corrected for errors encountered when trying to generate
the coverage map on source code inlined from external crates (including
macros and generics) by avoiding adding external DefIds to the coverage
map.
This made it possible to generate a coverage report including external
crates, but the external crate coverage was incomplete (did not include
coverage for the DefIds that were eliminated.
The root issue was that the coverage map was converting Span locations
to source file and locations, using the SourceMap for the current crate,
and this would not work for spans from external crates (compliled with a
different SourceMap).
The solution was to convert the Spans to filename and location during
MIR generation instead, so precompiled external crates would already
have the correct source code locations embedded in their MIR, when
imported into another crate.
@wesleywiser FYI
r? @tmandry
Update books
## reference
7 commits in b329ce37424874ad4db94f829a55807c6e21d2cb..c9b2736a059469043177e1e4ed41a55d7c63ac28
2020-07-20 08:54:08 -0700 to 2020-08-03 03:34:03 -0700
- Fix documented build output path. (rust-lang-nursery/reference#870)
- Update token usage table. (rust-lang-nursery/reference#868)
- Allow trait inner attributes (rust-lang-nursery/reference#864)
- patterns.md - add word "underscore" to _ paragraph (rust-lang-nursery/reference#865)
- Drive-by mention unsafe fn closure coercion (rust-lang-nursery/reference#802)
- grammar: Change "For awhile" to "For a while" (rust-lang-nursery/reference#857)
- Added Unpin to list of Auto Traits (rust-lang-nursery/reference#854)
## book
7 commits in a914f2c7e5cdb771fa465de142381a51c53b580e..363293c1c5ce9e84ea3935a5e29ce8624801208a
2020-07-21 09:20:05 -0500 to 2020-08-03 15:56:30 -0500
- replace commas with m-dashes to improve readability of chapter 4.1 (rust-lang/book#2419)
- Update TOML link to official website (rust-lang/book#2411)
- Add github repo link (rust-lang/book#2265)
- Remove the version number entirely so we can stop updating it
- Add link to the `Vec<T>` API documentation (rust-lang/book#2249)
- link to stdlib atomic docs (rust-lang/book#2361)
- mdbook version used is now 0.4.x (rust-lang/book#2410)
## rust-by-example
1 commits in 229c6945a26a53a751ffa4f9cb418388c00029d3..2e9271981adc32613365810f3428334c07095215
2020-07-06 10:13:15 -0300 to 2020-07-27 13:39:16 -0500
- Replaced "princess" with "royal" (rust-lang/rust-by-example#1363)
## embedded-book
1 commits in 94d9ea8460bcbbbfef1877b47cb930260b5849a7..b5256448a2a4c1bec68b93c0847066f92f2ff5a9
2020-07-05 14:17:40 +0000 to 2020-07-24 23:09:29 +0000
- Update c-with-rust.md (rust-embedded/book#256)
`run()` returns `Result<(), String>`. But on failure it always returns
an empty string, and then `wrap_return()` treats an empty string
specially, by not reporting the error.
It turns out we already have the `ErrorReported` type for this sort of
behaviour. This commit changes `run()` to use it.
rustdoc's `main()` immediately spawns a thread, M, with a large stack
(16MiB or 32MiB) on which it runs `main_args()`. `main_args()` does a
small amount of options processing and then calls
`setup_callbacks_and_run_in_default_thread_pool_with_globals()`, which
spawns it own thread, and M is not used further.
So, thread M seems unnecessary. However, it does serve a purpose: if the
options processing in `main_args()` panics, that panic is caught when M
is joined. So M can't simply be removed.
However, `main_options()`, which is called by `main_args()`, has a
`catch_fatal_errors()` call within it. We can move that call to `main()`
and change it to the very similar `catch_with_exit_code()`. With that in
place, M can be removed, and panics from options processing will still
be caught appropriately.
Even better, this makes rustdoc's `main()` match rustc's `main()`, which
also uses `catch_with_exit_code()`.
(Also note that the use of a 16MiB/32MiB stack was eliminated from rustc
in #55617.)
Remove the `--no-threads` workaround for wasm targets.
Remove `--no-threads` from the wasm-ld command-line, which was a
workaround for [an old bug] which was fixed in LLVM 9.0, and is
no longer needed.
Also, the `--no-threads` option has been [removed upstream].
[an old bug]: https://bugs.llvm.org/show_bug.cgi?id=41508
[removed upstream]: https://reviews.llvm.org/D76885
r? @alexcrichton
Remove `--no-threads` from the wasm-ld command-line, which was a
workaround for [an old bug] which was fixed in LLVM 9.0, and is
no longer needed.
Also, the `--no-threads` option has been [removed upstream].
[an old bug]: https://bugs.llvm.org/show_bug.cgi?id=41508
[removed upstream]: https://reviews.llvm.org/D76885
This commit adds some debug assertions to `ensure_monomorphic_enough`
which checks that unused generic parameters have been replaced with a
parameter.
Signed-off-by: David Wood <david@davidtw.co>
This commit skips encoding empty polymorphization results - while
polymorphization is disabled, this should be every polymorphization
result; but when polymorphization is re-enabled, this would help with
non-generic functions and those which do use all their parameters (most
functions).
Signed-off-by: David Wood <david@davidtw.co>
The prior PR corrected for errors encountered when trying to generate
the coverage map on source code inlined from external crates (including
macros and generics) by avoiding adding external DefIds to the coverage
map.
This made it possible to generate a coverage report including external
crates, but the external crate coverage was incomplete (did not include
coverage for the DefIds that were eliminated.
The root issue was that the coverage map was converting Span locations
to source file and locations, using the SourceMap for the current crate,
and this would not work for spans from external crates (compliled with a
different SourceMap).
The solution was to convert the Spans to filename and location during
MIR generation instead, so precompiled external crates would already
have the correct source code locations embedded in their MIR, when
imported into another crate.
This commit changes polymorphization to return a `FiniteBitSet<u32>`
rather than a `FiniteBitSet<u64>` because most functions do not use
anywhere near sixty-four generic parameters so keeping a `u64` around is
unnecessary in most cases.
Signed-off-by: David Wood <david@davidtw.co>
Remove two fields from `SubstFolder`.
They're only used in error messages printed if there's an internal
compiler error, and the cost of maintaining them is high enough to show
up in profiles.
r? @matthewjasper
ci: disable fast-fail on auto-fallible
The purpose of the auto-fallible job is to run builders that are likely to fail on CI without gating on them. Having fast-fail enabled there kinda defeats the purpose, as if one of them fails we can't monitor the outcome of the other ones.
This was prompted by the aarch64-gnu builder consistently failing due to a broken test, preventing us from seeing if the macOS spurious failure is fixed.
r? @Mark-Simulacrum
The purpose of the auto-fallible job is to run builders that are likely
to fail on CI without gating on them. Having fail-fast enabled there
kinda defeats the purpose, as if one of them fails we can't monitor the
outcome of the other ones.
This was prompted by the aarch64-gnu builder consistently failing due to
a broken test, preventing us from seeing if the macOS spurious failure
is fixed.
