rustdoc: remove hand-rolled isatty
This PR replaces bindings to the platform-specific isatty APIs with the `isatty` crate, as done elsewhere in the repository.
Update rayon and rustc-rayon
This updates rayon for various tools and rustc-rayon for the compiler's parallel mode.
- rayon v1.3.1 -> v1.5.1
- rayon-core v1.7.1 -> v1.9.1
- rustc-rayon v0.3.1 -> v0.3.2
- rustc-rayon-core v0.3.1 -> v0.3.2
... and indirectly, this updates all of crossbeam-* to their latest versions.
Fixes#92677 by removing crossbeam-queue, but there's still a lingering question about how tidy discovers "runtime" dependencies. None of this is truly in the standard library's dependency tree at all.
The rustc fork of rayon integrates with Cargo's jobserver to limit the
amount of parallelism. However, rustdoc's use case is concurrent I/O,
which is not CPU-heavy, so it should be able to use mainline rayon.
See this discussion [1] for more details.
[1]: https://github.com/rust-lang/rust/issues/90227#issuecomment-952468618
Note: I chose rayon 1.3.1 so that the rayon version used elsewhere in
the workspace does not change.
Since RFC 3052 soft deprecated the authors field anyway, hiding it from
crates.io, docs.rs, and making Cargo not add it by default, and it is
not generally up to date/useful information, we should remove it from
crates in this repo.
Replaces a format!() call in layout::render with a template
expansion. Introduces a `templates` field in SharedContext so parts
of rustdoc can share pre-rendered templates.
This currently builds in a copy of the single template available, like
with static files. However, future work can make this live-loadable with
a perma-unstable flag, to make rustdoc developers' work easier.
With the arrival of min const generics, many alt-vec libraries have
updated to use it in some way and arrayvec is no exception. Use the
latest with minor refactoring.
Also, rustc_workspace_hack is the only user of smallvec 0.6 in the
entire tree, so drop it.
Previously, logging at DEBUG or below would always be silenced, because
rustc compiles tracing with the `static_max_level_info` feature. That
makes sense for release artifacts, but not for developing rustdoc.
Instead, this compiles two different versions of tracing: one in the
release artifacts, distributed in the sysroot, and a new version
compiled by rustdoc. Since `rustc_driver` is always linked to the
version of sysroot, this copy/pastes `init_env_logging` into rustdoc.
The builds the second version of tracing unconditionally; see the code
for details on why.
* Reuse memory
* simplify `next_def_id`, avoid multiple hashing and unnecessary lookups
* remove `all_fake_def_ids`, use the global map instead (probably not a good step toward parallelization, though...)
* convert `add_deref_target` to iterative implementation
* use `ArrayVec` where we know the max number of elements
* minor touchups here and there
* avoid building temporary vectors that get appended to other vectors
At most places I may or may not be doing the compiler's job is this PR.
Thanks to marcusklaas' hard work in https://github.com/raphlinus/pulldown-cmark/pull/469, this fixes a lot of rustdoc bugs!
- Get rid of unnecessary `RefCell`
- Fix duplicate warnings for broken implicit reference link
- Remove unnecessary copy of links
The only change is that `expect_file` now uses path relative to the
current file (same as `include!`). Before, it used paths relative to
the workspace root, which makes no sense.
It's a unit-test in a sense that it only checks syntax highlighting.
However, the resulting HTML is written to disk and can be easily
inspected in the browser.
To update the test, run with `--bless` argument or set
`UPDATE_EXPEC=1` env var
This drops the parking_lot dependency; the ReentrantMutex type appeared
to be unused (at least, no compilation failures occurred).
This is technically a possible change in behavior of its users, as
lock() would wait on other threads releasing their guards, but since we
didn't actually remove any threading or such in this code, it appears
that we never used that behavior (the behavior change is only noticeable
if the type previously was used in two threads, in a single thread
ReentrantMutex is useless).
* Move fs::create_dir_all calls into DocFS to provide a clean
extension point if async extension there is needed.
* Convert callsites of create_dir_all to ensure_dir to reduce syscalls.
* Convert fs::write usage to DocFS.write
(which also removes a lot of try_err! usage for easier reading)
* Convert File::create calls to use Vec buffers and then DocFS.write
in order to consistently reduce syscalls as well, make
deferring to threads cleaner and avoid leaving dangling content if
writing to existing files....
* Convert OpenOptions usage similarly - I could find no discussion on
the use of create_new for that one output file vs all the other
files render creates, if link redirection attacks are a concern
DocFS will provide a good central point to introduce systematic
create_new usage. (fs::write/File::create is vulnerable to link
redirection attacks).
* DocFS::write defers to rayon for IO on Windows producing a modest
speedup: before this patch on my development workstation:
$ time cargo +mystg1 doc -p winapi:0.3.7
Documenting winapi v0.3.7
Finished dev [unoptimized + debuginfo] target(s) in 6m 11s
real 6m11.734s
Afterwards:
$ time cargo +mystg1 doc -p winapi:0.3.7
Compiling winapi v0.3.7
Documenting winapi v0.3.7
Finished dev [unoptimized + debuginfo] target(s) in 49.53s
real 0m49.643s
I haven't measured how much time is in the compilation logic vs in the
IO and outputting etc, but this takes it from frustating to tolerable
for me, at least for now.
Update parking_lot to 0.7
Unfortunately this'll dupe parking_lot until the data_structures crate
is published and be updated in rls in conjunction with crossbeam-channel
