This pulls out the implementations of most built-in lints into a
separate crate, to reduce edit-compile-test iteration times with
librustc_lint and increase parallelism. This should enable lints to be
refactored, added and deleted much more easily as it slashes the
edit-compile cycle to get a minimal working compiler to test with (`make
rustc-stage1`) from
librustc -> librustc_typeck -> ... -> librustc_driver ->
libcore -> ... -> libstd
to
librustc_lint -> librustc_driver -> libcore -> ... libstd
which is significantly faster, mainly due to avoiding the librustc build
itself.
The intention would be to move as much as possible of the infrastructure
into the crate too, but the plumbing is deeply intertwined with librustc
itself at the moment. Also, there are lints for which diagnostics are
registered directly in the compiler code, not in their own crate
traversal, and their definitions have to remain in librustc.
This is a [breaking-change] for direct users of the compiler APIs:
callers of `rustc::session::build_session` or
`rustc::session::build_session_` need to manually call
`rustc_lint::register_builtins` on their return value.
This should make #22206 easier.
After making `rustc` fail on errors at a stop point, like `-Z parse-only`, in #22117, the files in this PR also fail during the parse stage and should be moved as well. Sorry for spliting this move up in two PRs.
- Now "make check-stage2-T-aarch64-linux-android-H-x86_64-unknown-linux-gnu" works (#21773)
- Fix & enable debuginfo tests for android (#10381)
- Fix & enable more tests for android (both for arm/aarch64)
- Enable many already-pass tests on android (both for arm/aarch64)
`cp -a` is a GNU extension. Use an alternate combinaison of POSIX options
(`-PRp`) that do nearly the same.
The difference is `-a` will preserve context, links and xattr attributes,
whereas `-p` not. But as we use it only for copy a file, there is no
difference in the current context.
This is a patch for #22291.
PLEASE_BENCH=1 adds --bench to the arguments passed to the executable to be tested. At the moment, compiletest does not accept a --bench argument, because it is not needed for any test in src/test/, even the tests in src/test/bench do not use #[bench].
I have updated the makefile to only add the --bench flag for crate tests. I do not think that changing compiletest add --bench to the run arguments of all compile tests makes sense, because it would mess up tests which check command line arguments. Also the bench option can be added as comment in a compile test as well.
Highlights:
* Adds an 'uninstall.sh' script to `/usr/local/lib/rustlib/uninstall.sh`, the path to which is printed during installation.
* Components can be deselected during install, like `install.sh --without=rust-docs`.
* Components can be listed with `install.sh --list-components`.
* Vastly reduces spew during install (but supporting a `--verbose` option).
Typicall install run looks like:
```
brian@brianX1:~/dev/multirust⟫ sudo ./install.sh
[sudo] password for brian:
install: creating uninstall script at /usr/local/lib/rustlib/uninstall.sh
install: installing component 'rustc'
install: installing component 'cargo'
install: installing component 'rust-docs'
Rust is ready to roll.
```
Needs to be merged right before corresponding PRs to cargo and rust-packaging.
Fixes https://github.com/rust-lang/rust/issues/21117
Fixes https://github.com/rust-lang/rust/issues/20283
`cp -a` is a GNU extension. Use an alternate combinaison of POSIX options
(`-PRp`) that do nearly the same.
The difference is `-a` will preserve context, links and xattr attributes,
whereas `-p` not. But as we use it only for copy a file, there is no
difference in the current context.
This restructures tidy.py to walk the tree itself,
and improves performance considerably by not loading entire
files into buffers for licenseck.
Splits build rules into 'tidy', 'tidy-basic', 'tidy-binaries',
'tidy-errors', 'tidy-features'.
This adds a new lexer/parser combo for the entire Rust language can be generated with with flex and bison, taken from my project at https://github.com/bleibig/rust-grammar. There is also a testing script that runs the generated parser with all *.rs files in the repository (except for tests in compile-fail or ones that marked as "ignore-test" or "ignore-lexer-test"). If you have flex and bison installed, you can run these tests using the new "check-grammar" make target.
This does not depend on or interact with the existing testing code in the grammar, which only provides and tests a lexer specification.
OS X users should take note that the version of bison that comes with the Xcode toolchain (2.3) is too old to work with this grammar, they need to download and install version 3.0 or later.
The parser builds up an S-expression-based AST, which can be displayed by giving the "-v" argument to parser-lalr (normally it only gives output on error). It is only a rough approximation of what is parsed and doesn't capture every detail and nuance of the program.
Hopefully this should be sufficient for issue #2234, or at least a good starting point.
The regex library was largely used for non-critical aspects of the compiler and
various external tooling. The library at this point is duplicated with its
out-of-tree counterpart and as such imposes a bit of a maintenance overhead as
well as compile time hit for the compiler itself.
The last major user of the regex library is the libtest library, using regexes
for filters when running tests. This removal means that the filtering has gone
back to substring matching rather than using regexes.
Initial support for aarch64-linux-android (#18920)
- Add new configuration files
- Modify some options to compile & link succesfully.
(PIE, disable tls on jemalloc, modify some external function linkage, ..)
- To build, refer to https://github.com/rust-lang/rust/wiki/Doc-building-for-android.
(tested with platform=21 and toolchain=aarch64-linux-android-4.9)
This does the bare minimum to make registration of error codes work again. After this patch, every call to `span_err!` with an error code gets that error code validated against a list in that crate and a new tidy script `errorck.py` validates that no error codes are duplicated globally.
There are further improvements to be made yet, detailed in #19624.
r? @nikomatsakis
Removed use of unused LDPATH variable on Windows as is done for other platforms, and added GCC flag to ensure MINGW's ANSI compatible STDIO functions are used wherever available (required by jemalloc).
Without these changes it ends up setting the PATH twice, and the second time the PATH begins with `:` which is invalid. Also the regular msvcrt printf-like functions would be used which don't understand stuff like %hhd and %z which jemalloc uses.
This change ought not to make any difference to the output but it fixes the build process for me since at least my build environment couldn't handle that broken path caused by LDPATH being empty.
