coverage bug fixes and optimization support
Adjusted LLVM codegen for code compiled with `-Zinstrument-coverage` to
address multiple, somewhat related issues.
Fixed a significant flaw in prior coverage solution: Every counter
generated a new counter variable, but there should have only been one
counter variable per function. This appears to have bloated .profraw
files significantly. (For a small program, it increased the size by
about 40%. I have not tested large programs, but there is anecdotal
evidence that profraw files were way too large. This is a good fix,
regardless, but hopefully it also addresses related issues.
Fixes: #82144
Invalid LLVM coverage data produced when compiled with -C opt-level=1
Existing tests now work up to at least `opt-level=3`. This required a
detailed analysis of the LLVM IR, comparisons with Clang C++ LLVM IR
when compiled with coverage, and a lot of trial and error with codegen
adjustments.
The biggest hurdle was figuring out how to continue to support coverage
results for unused functions and generics. Rust's coverage results have
three advantages over Clang's coverage results:
1. Rust's coverage map does not include any overlapping code regions,
making coverage counting unambiguous.
2. Rust generates coverage results (showing zero counts) for all unused
functions, including generics. (Clang does not generate coverage for
uninstantiated template functions.)
3. Rust's unused functions produce minimal stubbed functions in LLVM IR,
sufficient for including in the coverage results; while Clang must
generate the complete LLVM IR for each unused function, even though
it will never be called.
This PR removes the previous hack of attempting to inject coverage into
some other existing function instance, and generates dedicated instances
for each unused function. This change, and a few other adjustments
(similar to what is required for `-C link-dead-code`, but with lower
impact), makes it possible to support LLVM optimizations.
Fixes: #79651
Coverage report: "Unexecuted instantiation:..." for a generic function
from multiple crates
Fixed by removing the aforementioned hack. Some "Unexecuted
instantiation" notices are unavoidable, as explained in the
`used_crate.rs` test, but `-Zinstrument-coverage` has new options to
back off support for either unused generics, or all unused functions,
which avoids the notice, at the cost of less coverage of unused
functions.
Fixes: #82875
Invalid LLVM coverage data produced with crate brotli_decompressor
Fixed by disabling the LLVM function attribute that forces inlining, if
`-Z instrument-coverage` is enabled. This attribute is applied to
Rust functions with `#[inline(always)], and in some cases, the forced
inlining breaks coverage instrumentation and reports.
FYI: `@wesleywiser`
r? `@tmandry`
Add documentation for rustdoc-gui tests
I think a bit of documentation doesn't hurt in this case considering how "out of the ordinary" this is.
r? ``@jyn514``
Remove Option::{unwrap_none, expect_none}.
This removes `Option::unwrap_none` and `Option::expect_none` since we're not going to stabilize them, see https://github.com/rust-lang/rust/issues/62633.
Closes#62633
Fixes#80691
When we evaluate a trait predicate, we convert an
`EvaluatedToOk` result to `EvaluatedToOkModuloRegions` if we erased any
regions. We cache the result under a region-erased 'freshened'
predicate, so `EvaluatedToOk` may not be correct for other predicates
that have the same cache key.
2229 migration: Don't try resolve regions before writeback
In the analysis use `resolve_vars_if_possible` instead of `fully_resolve`,
because we might not have performed regionck yet.
Fixes: #83176
r? `@nikomatsakis`
implement `feature(const_generics_defaults)`
Implements const generics defaults `struct Example<const N: usize=3>`, as well as a query for getting the default of a given const-parameter's def id. There are some remaining FIXME's but they were specified as not blocking for merging this PR. This also puts the defaults behind the unstable feature gate `#![feature(const_generics_defaults)]`.
~~This currently creates a field which is always false on `GenericParamDefKind` for future use when
consts are permitted to have defaults. I'm not sure if this is exactly what is best for adding default parameters, but I mimicked the style of type defaults, so hopefully this is ok.~~
r? `@lcnr`
Remove unnecessary `Option` wrapping around `Crate.module`
I'm wondering if it was originally there so that we could `take` the
module which enables `after_krate` to take an `&Crate`. However, the two
impls of `after_krate` only use `Crate.name`, so we can pass just the
name instead.
Slight visual improvements to warning boxes in the docs
First I noticed that sometimes the thumbs-down emoji in the docs is hard to see and hard to look at because the yellow emoji color and the color of the box below are so bright. Especially if you look at the screen late at night you can notice it. I thought I should change that so I added a black outline around the emoji. It works using the [`text-shadow`](https://developer.mozilla.org/en-US/docs/Web/CSS/text-shadow) property. It may be a bit hacky but it seems to work well and browser compatibility looks pretty good too: [browser compatibility](https://developer.mozilla.org/en-US/docs/Web/CSS/text-shadow#browser_compatibility).
For consistency the microscope has the black border too.
Alternatively I had `drop-shadow(0px 0px 1px black);` in mind but its [browser compatibility](https://developer.mozilla.org/en-US/docs/Web/CSS/filter-function/drop-shadow()#browser_compatibility) doesn't look as good and the blurry shadow probably doesn't look as good either.
Then, I thought that now that I'm at it I could also try changing the purple color to a color you would rather expect to see for deprecation: red. For the red I've taken the blue and reused it as a foundation and moved it to the red color spectrum.
But then I thought that the purple color could still be reused for something else: for the boxes that tell you about portability (e.g. _only supported on Unix_). These are currently blue.
I think blue doesn't really represent danger like it should. Not being cross-platform represents a danger because if you want to compile for a different platform, your code may not compile anymore. Blue looks too friendly and is in my opinion more suitable for a box containing general information like for instance "This is available since 1.0.0". None of the current three box types (unstable, deprecated and portability) are that.
I think purple is a better fit for it because it's kind of in the middle between "use it" and "don't use it". Deprecated is definitely "don't use it". To illustrate this better, here's a color spectrum:
Blue = friendly, "use it".
Red = danger, "don't use it".
And the purple in the middle (the color that the portability box now has) probably represents "use it if you have to", so it's not entirely friendly and not entirely a danger. That is why I think it fits.
However I made one change to that existing purple: I made the outer color a bit brighter because it's outstandingly dark compared to the other outer colors of the other boxes.
This is all subjective but in my opinion it looks nicer. At first you might need to get used to it though. Notice the box colors and the black outlines around the emoji shapes:
Allow not emitting `uwtable` on Android
`uwtable` is marked as required on Android, so it can't be disabled via `-C force-unwind-tables=no`. However, I found that the reason it's marked as required was to resolve a [backtrace issue in Gecko](https://github.com/rust-lang/rust/issues/49867), and I haven't find any other reasons that make it required ([yet](https://rust-lang.zulipchat.com/#narrow/stream/122651-general/topic/Unwind.20tables.20are.20strictly.20required.20on.20Windows.20and.20Android)). Therefore, I assume it's safe to turn it off if a (nice) backtrace is not needed, and submit this PR to allow `-C force-unwind-tables=no` when targeting Android.
Note that I haven't tested this change on Android as I don't have an Android environment for testing.
Only enable assert_dep_graph when query-dep-graph is enabled.
This is a debugging option. The only effect should be on rustc tests.
r? ``@michaelwoerister``
Sidebar trait items order
We were actually sorting `Symbol` and not `String`, creating a completely invalid sort result. I added a test to prevent regressions.
r? ``@jyn514``
Rename `source` to `span` and `span` to `source`
- Rename `clean::Item.source` to `span`
- Rename `clean::Span::span()` to `clean::Span::inner()`
- Rename `rustdoc_json_types::Item.source` to `span`
- rustdoc-json: Rename `Import.span` to `Import.source`
*See also the [discussion on Zulip][z] (this is a bit more than discussed in
that conversation, but all the changes are related).*
r? `@jyn514`
[z]: https://rust-lang.zulipchat.com/#narrow/stream/182449-t-compiler.2Fhelp/topic/get.20span.20of.20file.20from.20name/near/229603729
I'm wondering if it was originally there so that we could `take` the
module which enables `after_krate` to take an `&Crate`. However, the two
impls of `after_krate` only use `Crate.name`, so we can pass just the
name instead.
This currently creates a field which is always false on GenericParamDefKind for future use when
consts are permitted to have defaults
Update const_generics:default locations
Previously just ignored them, now actually do something about them.
Fix using type check instead of value
Add parsing
This adds all the necessary changes to lower const-generics defaults from parsing.
Change P<Expr> to AnonConst
This matches the arguments passed to instantiations of const generics, and makes it specific to
just anonymous constants.
Attempt to fix lowering bugs
Now that we record the crate's name in its `clean::Item`, pushing the
crate name onto the `stack` causes duplicate paths. E.g., the URL
generated for the path `::foo::bar::baz` would be something like
../foo/foo/bar/baz
With this commit, the URL is corrected to
../foo/bar/baz
rustdoc: Replace pair of `Option`s with an enum
They are never both `None` or both `Some`, so it makes more sense to use
an enum so that we "make impossible states impossible".
Remove theme.js file
Fixes#82616.
The first commit moves the `theme.js` file into `main.js`, which requires to also run a small `.replace` on the `main.js` content.
The second commit is just a small cleanup to centralize DOM ids.
Since it removes a file from rustdoc output: cc `@rust-lang/docs-rs`
cc `@jsha`
r? `@jyn514`
