`FileSearch::search()` traverses one or more directories. For each
directory it generates a `Vec<PathBuf>` containing one element per file
in that directory.
In some benchmarks this occurs enough that the allocations done for the
`PathBuf`s are significant, and in practice a small number of
directories are being traversed over and over again. For example, when
compiling the `tokio-webpush-simple` benchmark, two directories are
traversed 58 times each. Each of these directories have more than 100
files.
This commit changes things so that all the `Vec<PathBuf>`s that will be
needed by a `Session` are precomputed when that `Session` is created;
they are stored in `SearchPath`. `FileSearch` gets a reference to the
necessary `SearchPath`s. This reduces instruction counts on several
benchmarks by 1--5%.
The commit also removes the barely-used `visited_dirs` hash in
`for_each_lib_searchPath`. It only detects if `tlib_path` is the same as
one of the previously seen paths, which is unlikely.
Instead of maybe storing its own sysroot and maybe deferring to the one
in `Session::opts`, just clone the latter when necessary so one is
always directly available. This removes the need for the getter.
This shrinks `Option<Symbol>` from 8 bytes to 4 bytes, which shrinks
`Token` from 24 bytes to 16 bytes. This reduces instruction counts by up
to 1% across a range of benchmarks.
Test with gdb8.2 and add debuginfo printing function call test
As far as I can see, `print function()` is not tested. It is important feature for debugging.
Initially, #50233 accidentally changed the capacity of empty ZST. This
was pointed out during code review. This commit adds a test to prevent
capacity of ZST vectors from accidentally changing to prevent that
from happening again.
libtest: Use deterministic HashMap, avoid spawning thread if there is no concurrency
It seems desirable to make a test and bench runner deterministic, which this achieves by using a deterministic hasher. Also, we we only have 1 thread, we don't bother spawning one and just use the main thread.
The motivation for this is to be able to run the test harness in miri, where we can neither access the OS RNG, nor spawn threads.
This builder is not really the correct place to put this, but it
definitely has the time budget and checking this tool builds on just one
platform is more than sufficient.
Update Cargo submodule and its dependencies
Hopefully just another routine update!
So far this starts to enable the `std::arch` in stage0 builds of rustc.
This means that we may need stage0/not(stage0) in stdsimd itself, but
more and more code is starting to use `std::arch` so I think it's time
to start shifting the balance of work here.
This commit improves the calculation of code spans for intra-doc
resolution failures. All sugared doc comments should now have the
correct spans, including those where the comment is longer than the
docs.
It also fixes an issue where the spans were calculated incorrectly for
certain unsugared doc comments. The diagnostic will now always use the
span of the attributes, as originally intended.
Fixes#55964.
Hopefully just another routine update!
So far this starts to enable the `std::arch` in stage0 builds of rustc.
This means that we may need stage0/not(stage0) in stdsimd itself, but
more and more code is starting to use `std::arch` so I think it's time
to start shifting the balance of work here.
Rollup of 5 pull requests
Successful merges:
- #56491 (emit error with span for empty asserts)
- #56633 (Fix right arrow size for crate filter)
- #56641 (fix span for invalid number of parameters in trait method)
- #56656 (Fix typo)
- #56661 (Add regression test for ICE)
Failed merges:
r? @ghost
Add regression test for ICE
Fixes#55846 with a minimal (or as best as I can manage) test case. I tested this against 1.30.0 manually to confirm it crashes.
The issue seemed to have something to do with associated types. It's possible someone with more knowledge can shrink the test case down further, or make it clearer.
