Puts implementations in bins hashed by the fast-reject key, and
only looks up the relevant impls, reducing O(n^2)-ishness
Before: 688.92user 5.08system 8:56.70elapsed 129%CPU (0avgtext+0avgdata 1208164maxresident)k, LLVM 379.142s
After: 637.78user 5.11system 8:17.48elapsed 129%CPU (0avgtext+0avgdata 1201448maxresident)k LLVM 375.552s
Performance increase is +7%-ish
The former stopped making sense when we started interning substs and made
TraitRef a 2-word copy type, and I'm moving the latter into an arena as
they live as long as the type context.
I've been working on improving the diagnostic registration system so that it can:
* Check uniqueness of error codes *across the whole compiler*. The current method using `errorck.py` is prone to failure as it relies on simple text search - I found that it breaks when referencing an error's ident within a string (e.g. `"See also E0303"`).
* Provide JSON output of error metadata, to eventually facilitate HTML output, as well as tracking of which errors need descriptions. The current schema is:
```
<error code>: {
"description": <long description>,
"use_site": {
"filename": <filename where error is used>,
"line": <line in file where error is used>
}
}
```
[Here's][metadata-dump] a pretty-printed sample dump for `librustc`.
One thing to note is that I had to move the diagnostics arrays out of the diagnostics modules. I really wanted to be able to capture error usage information, which only becomes available as a crate is compiled. Hence all invocations of `__build_diagnostics_array!` have been moved to the ends of their respective `lib.rs` files. I tried to avoid moving the array by making a plugin that expands to nothing but couldn't invoke it in item position and gave up on hackily generating a fake item. I also briefly considered using a lint, but it seemed like it would impossible to get access to the data stored in the thread-local storage.
The next step will be to generate a web page that lists each error with its rendered description and use site. Simple mapping and filtering of the metadata files also allows us to work out which error numbers are absent, which errors are unused and which need descriptions.
[metadata-dump]: https://gist.github.com/michaelsproul/3246846ff1bea71bd049
Changes made include adding missing punctuation, adding missing words, and converting uses of "Gets" to "Returns" in libstd/net/addr.rs to make it more consistent with the other documentation.
Fixes#24925.
This is OK to do given:
- PIE is supported on Android starting with API 16.
- The bots are running API 18.
- API < 16 now has a 12.5% market share[0] as of 2015-04-29.
Closes#17437.
[0] https://developer.android.com/about/dashboards/index.html
r? @alexcrichton
Add `-g` (to testcase) that I should have included in PR #24932.
Note it is safe, with respect to autobuilds, to land before #24945.
(In other words, landing this sooner won't break things for anyone any
worse than they were already broken, since there are *other* tests
that also add `-g` to their flags via `compile-flags: -g`.)
Fixes for -g handling
First:
* decouples our handling of `-g` for the test suite from our handling of `-g` for the rest of the compiler/stdlib building.
* Namely, if you do `--enable-debug` or `--enable-debuginfo`, that should only affect `rustc` and the standard library crates; the tests should all continue to compile without `-g` unless:
* you pass `--enable-debuginfo-tests`, or
* the test itself requests the `-g` option (e.g. via a `// compile-flags: -g` embedded comment).
Second:
* Makes `rustc` more flexible in that it now accepts multiple occurrences of `-g -g`
* (as a drive-by, I gave `-O` the same treatment: multiple occurrences of `-O` are treated as synonymous as a single occurrence of `-O`.
Fix#24937
Since #24783, the style guidelines recommend that unit tests should live in a submodule `tests` rather than `test` to not clash with the possible use of libtest. This is especially important for benchmark tests as they require libtest. Fixes#24923.
`call_once` guarantees that there is a happens-before relationship between its closure and code following it via the sequentially consistent atomic store/loads of `self.cnt`.
This error indicates that a constant references itself.
All constants need to resolve to a value in an acyclic manner.
For example, neither of the following can be sensibly compiled:
```
const X: u32 = X;
```
```
const X: u32 = Y;
const Y: u32 = X;
```
These are useful when you want to catch the signals, like when you're making a kernel, or if you just don't want the overhead. (I don't know if there are any of the second kind of people, I don't think it's a good idea, but hey, choice is good).
Currently, LLVM lowers a cttz8 on x86_64 to these instructions:
```asm
movzbl %dil, %eax
bsfl %eax, %eax
movl $32, %ecx
cmovnel %eax, %ecx
cmpl $32, %ecx
movl $8, %eax
cmovnel %ecx, %eax
```
To improve the codegen, we can zero extend the 8 bit integer, then set
bit 8 and perform a cttz operation on the extended value. That way
there's no conditional operation involved at all.
This was discovered by this benchmark: https://github.com/Kimundi/long_strings_without_repeats
Timings on my box with the current nightly:
```
running 4 tests
test bench_cpp_naive_big ... bench: 5479222 ns/iter (+/- 254222)
test bench_noop_big ... bench: 571405 ns/iter (+/- 111950)
test bench_rust_naive_big ... bench: 7798102 ns/iter (+/- 148841)
test bench_rust_unsafe_big ... bench: 6606488 ns/iter (+/- 67529)
```
Timings with the patch applied:
```
running 4 tests
test bench_cpp_naive_big ... bench: 5470944 ns/iter (+/- 7109)
test bench_noop_big ... bench: 568944 ns/iter (+/- 6895)
test bench_rust_naive_big ... bench: 6795901 ns/iter (+/- 43806)
test bench_rust_unsafe_big ... bench: 5584879 ns/iter (+/- 5291)
```
This commit is an implementation of [RFC 1044][rfc] which adds additional
surface area to the `std::fs` module. All new APIs are `#[unstable]` behind
assorted feature names for each one.
[rfc]: https://github.com/rust-lang/rfcs/pull/1044
The new APIs added are:
* `fs::canonicalize` - bindings to `realpath` on unix and
`GetFinalPathNameByHandle` on windows.
* `fs::symlink_metadata` - similar to `lstat` on unix
* `fs::FileType` and accessor methods as `is_{file,dir,symlink}`
* `fs::Metadata::file_type` - accessor for the raw file type
* `fs::DirEntry::metadata` - acquisition of metadata which is free on Windows
but requires a syscall on unix.
* `fs::DirEntry::file_type` - access the file type which may not require a
syscall on most platforms.
* `fs::DirEntry::file_name` - access just the file name without leading
components.
* `fs::PathExt::symlink_metadata` - convenience method for the top-level
function.
* `fs::PathExt::canonicalize` - convenience method for the top-level
function.
* `fs::PathExt::read_link` - convenience method for the top-level
function.
* `fs::PathExt::read_dir` - convenience method for the top-level
function.
* `std::os::raw` - type definitions for raw OS/C types available on all
platforms.
* `std::os::$platform` - new modules have been added for all currently supported
platforms (e.g. those more specific than just `unix`).
* `std::os::$platform::raw` - platform-specific type definitions. These modules
are populated with the bare essentials necessary for lowing I/O types into
their raw representations, and currently largely consist of the `stat`
definition for unix platforms.
This commit also deprecates `Metadata::{modified, accessed}` in favor of
inspecting the raw representations via the lowering methods of `Metadata`.
Closes https://github.com/rust-lang/rust/issues/24796
