also export metrics from librustdoc
Addresses the issue mentioned here: [#t-docs-rs > metrics intitiative @ 💬](https://rust-lang.zulipchat.com/#narrow/channel/356853-t-docs-rs/topic/metrics.20intitiative/near/515714331)
The previous implementation only emitted metrics from rustc, but it turns out running `cargo doc` only calls `rustc` for dependencies, and not for the root crate being documented. We are planning to gather a sample dataset from docs.rs ci via `cargo doc` so as things stood this would not emit any metrics for any of the crates themselves that were published.
This change adds the same logic from `rustc_driver_impl` to `librustdoc` to also dump metrics at the end of its execution if they are enabled.
Note: The hash's generated by librustdoc will likely be completely different from the ones generated by rustc. This is because rustc is actually doing the various passes needed to fully calculate the stable version hash. My understanding of how rustdoc works is that the hashes generated will be working with partial information due to it only doing the work required to generate docs. The hashes will still be unique per crate and will work for the purposes of the metrics proof of concept, it would not be possible to correlate metrics generated by rustdoc with those generated by rustc for the same crate. This is fine for the purposes of the PoC but a future full implementation of metrics may want to address this issue.
rustdoc: Replace unstable flag `--doctest-compilation-args` with a simpler one: `--doctest-build-arg`
Tracking issue: https://github.com/rust-lang/rust/issues/134172.
Context: https://github.com/rust-lang/rust/pull/137096#issuecomment-2776318800
Yeets the ad hoc shell-like lexer for 'nested' program arguments.
No FCP necessary since the flag is unstable.
I've chosen to replace `compilation` with `build` because it's shorter (you now need to pass it multiple times in order to pass many arguments to the doctest compiler, so it matters a bit) and since I prefer it esthetically.
**Issue**: Even though we don't process the argument passed to `--doctest-build-arg`, we end up passing it via an argument file (`rustc `@argfile`)` which delimits arguments by line break (LF or CRLF, [via](https://doc.rust-lang.org/rustc/command-line-arguments.html#path-load-command-line-flags-from-a-path)) meaning ultimately the arguments still get split which is unfortunate. Still, I think this change is an improvement over the status quo.
I'll update the tracking issue if/once this PR merges. I'll also add the (CR)LF issue to 'unresolved question'.
r? GuillaumeGomez
r? notriddle
This cleans up `unwrap.rs`:
- use interned symbols instead of strings
- update names to reflect the current implementation
- replaced a cascaded `if` by a shorter `match`
changelog: none
This case can't actually happen yet (other than via a testing flag), because
currently all of a function's spans must belong to the same file and expansion.
But this will be an important edge case when adding expansion region support.
Where host compiler runtime libs are properly configured, instead of raw
`RUSTC`/`RUSTDOC` commands.
Co-authored-by: Jesus Checa Hidalgo <jchecahi@redhat.com>
In rare cases, the test may need access to the underlying
`std::process::Command` (e.g. for non-trivial process spawning).
Co-authored-by: Jesus Checa Hidalgo <jchecahi@redhat.com>
Previously `-Zprint-mono-items` would override the mono item collection
strategy. When debugging one doesn't want to change the behaviour, so
this was counter productive. Additionally, the produced behaviour was
artificial and might never arise without using the option in the first
place (`-Zprint-mono-items=eager` without `-Clink-dead-code`). Finally,
the option was incorrectly marked as `UNTRACKED`.
Resolve those issues, by turning `-Zprint-mono-items` into a boolean
flag that prints results of mono item collection without changing the
behaviour of mono item collection.
For codegen-units test incorporate `-Zprint-mono-items` flag directly
into compiletest tool.
Test changes are mechanical. `-Zprint-mono-items=lazy` was removed
without additional changes, and `-Zprint-mono-items=eager` was turned
into `-Clink-dead-code`. Linking dead code disables internalization, so
tests have been updated accordingly.
Revert "Rollup merge of #129343 - estebank:time-version, r=jieyouxu"
This reverts commit 26f75a65d7, reversing
changes made to 2572e0e8c9.
Imports are modified to fix merge conflicts and remove unused ones.
Enable non-leaf Frame Pointers for Arm64 Windows
Microsoft recommends enabling frame pointers for Arm64 Windows as it enables fast stack walking, from <https://learn.microsoft.com/en-us/cpp/build/arm64-windows-abi-conventions?view=msvc-170#integer-registers>:
> The frame pointer (x29) is required for compatibility with fast stack walking used by ETW and other services. It must point to the previous {x29, x30} pair on the stack.
I'm setting this to "non-leaf" as leaf functions shouldn't be spilling registers and so won't touch the frame pointer.
Use span before macro expansion in lint for-loops-over-falibles
Fixes#140747
I think there are going to be a lot of cases where macros are expanded in the compiler resulting in span offsets, and I'd like to know how that's typically handled. Does it have to be handled specially every time?
All uses have been removed. And it's nonsensical: an identifier by
definition has at least one char.
The commits adds an is-non-empty assertion to `Ident::new` to enforce
this, and converts some `Ident` constructions to use `Ident::new`.
Adding the assertion requires making `Ident::new` and
`Ident::with_dummy_span` non-const, which is no great loss.
The commit amends a couple of places that do path splitting to ensure no
empty identifiers are created.
