compiletest normalization: preserve non-JSON lines such as ICEs
Currently, every non-JSON line from stderr gets normalized away when compiletest normalizes the output. In particular, ICEs get normalized to the empty output. That does not seem desirable, so this changes normalization to preserve non-JSON lines instead.
Also see https://github.com/laumann/compiletest-rs/issues/169: because of that bug, Miri currently *looks* green in the toolstate, but some tests ICE. That same bug is likely no longer present in latest compiletest because the error code gets checked separately, but it still seems like a good idea to also make sure that ICEs are considered stderr output:
This change found an accidental user-visible `error!` in CTFE validation (fixed), and a non-deterministic panic when there are two `main` symbols (not fixed, no idea where this comes from). Both got missed before because non-JSON output got ignored.
rustdoc: use --static-root-path for settings.js
At the time i was writing https://github.com/rust-lang/docs.rs/pull/332, i noticed that the `settings.js` file that was being loaded was not being loaded from the `--static-root-path`. This PR fixes that so that users on docs.rs can effectively cache this file.
ci: use a custom android sdk manager with pinning and mirroring
Google's own sdkmanager has two issues that make it unsuitable for us:
* Mirroring has to be done manually, which is annoying because we need to figure out on our own all the URLs to copy (I couldn't find any documentation when building this PR, had to use mitmproxy).
* There is no support for pinning, which means an update on Google's side can break our CI, as it happened multiple times.
This PR replaces all our usage of sdkmanager with a custom Python script which mimics its behavior, but with the two issues fixes.
sdkmanager's logic for installing packages is thankfully very simple: the package name (like `system-images;android-18;default;armeabi-v7a`) is the directory where the package should live (with `;` replaced with `/`), so to install a package we only need to extract its contents in the right directory.
r? @alexcrichton
cc @kennytm
fixes https://github.com/rust-lang/rust/issues/59778
rustdoc: Remove default keyword from re-exported trait methods
Fixes#59977
r? @QuietMisdreavus
As this fixes a stable to beta regression, could it be backported?
bump stdsimd; make intra_doc_link_resolution_failure an error again; make lints more consistent
I made `intra_doc_link_resolution_failure` warn so that it would properly respect `deny-warnings = false` in `config.toml`. `#[warn]` still become errors with `-D warnings` so I thought this was fine.
Turns out however that we don't pass `-D warnings` when running rustdoc, so for additional rustdoc-lints we need to set them to `deny`.
Also sue the opportunity to make the lint flags more consistent between libcore, liballoc, libstd.
Cc @gnzlbg for the *big* stdsimd update.
Uplift `get_def_path` from Clippy
cc rust-lang/rust-clippy#3926
cc #59738
This uplifts `get_def_path` from Clippy. This is a follow up on the
implementation of internal lints: #59316
The internal lint implementation also copied the implementation of the
`AbsolutePathPrinter`. To get rid of this code duplication this also
uplifts the `get_def_path` function from Clippy.
This also renames `match_path` to `match_def_path`, as it was originally
named in Clippy.
r? @Manishearth
Reduce the `DepNode` pre-allocation ratio.
A code size of increase of 15% is overly generous. 2% is more realistic.
This change reduces peak memory size by 20+ MiB on some workloads.
r? @Zoxc
