When performing a "fat" LTO the compiler has a whole mess of codegen
units that it links together. To do this it needs to select one module
as a "base" module and then link everything else into this module.
Previously LTO passes assume that there's at least one module in-memory
to link into, but nowadays that's not always true! With incremental
compilation modules may actually largely be cached and it may be
possible that there's no in-memory modules to work with.
This commit updates the logic of the LTO backend to handle modules a bit
more uniformly during a fat LTO. This commit immediately splits them
into two lists, one serialized and one in-memory. The in-memory list is
then searched for the largest module and failing that we simply
deserialize the first serialized module and link into that. This
refactoring avoids juggling three lists, two of which are serialized
modules and one of which is half serialized and half in-memory.
Closes#63349
When deny-warnings is not specified or set to true, the behaviour is the same as before.
When deny-warnings is set to false, warnings are now allowed
Fixes#63911
Signed-off-by: Marc-Antoine Perennou <Marc-Antoine@Perennou.com>
Rollup of 4 pull requests
Successful merges:
- #62600 (libtest: add --show-output flag to print stdout of successful tests)
- #63698 (Fixed floating point issue with asinh function)
- #63761 (Propagate spans and attributes from proc macro definitions)
- #63917 (Error when generator trait is not found)
Failed merges:
r? @ghost
Propagate spans and attributes from proc macro definitions
Thanks to https://github.com/rust-lang/rust/pull/63269 we now have spans and attributes from proc macro definitions available in metadata.
However, that PR didn't actually put them into use! This PR finishes that work.
Attributes `rustc_macro_transparency`, `allow_internal_unstable`, `allow_internal_unsafe`, `local_inner_macros`, `rustc_builtin_macro`, `stable`, `unstable`, `rustc_deprecated`, `deprecated` now have effect when applied to proc macro definition functions.
From those attributes only `deprecated` is both stable and supposed to be used in new code.
(`#![staged_api]` still cannot be used in proc macro crates for unrelated reasons though.)
`Span::def_site` from the proc macro API now returns the correct location of the proc macro definition.
Also, I made a mistake in https://github.com/rust-lang/rust/pull/63269#discussion_r312702919, loaded proc macros didn't actually use the resolver cache.
This PR fixes the caching issue, now proc macros go through the `Resolver::macro_map` cache as well.
(Also, the first commit turns `proc_macro::quote` into a regular built-in macro to reduce the number of places where `SyntaxExtension`s need to be manually created.)
Fixed floating point issue with asinh function
This should fixes#63271 , in which `asinh(-0.0)` returns `0.0` instead of `-0.0`.
according to @nagisa
>
>
> IEEE-754 (2008), section 9.2.1:
>
> > For the functions expm1, exp2m1, exp10m1, logp1, log2p1, log10p1, sin, tan, sinPi, atanPi, asin, atan, sinh, tanh, asinh, and atanh, f(+0) is +0 and f(−0) is −0 with no exception.
>
> and
>
> > sinh(±∞) and asinh(±∞) are ±∞ with no exception.
After ensuring that the function `asinh` is the only function affected (functions like `sin`, `sinh` are all based on `cmath` library or `llvm` intrinsics), and that `atanh` always gives the correct result. The only function to modify is `asinh`.
libtest: add --show-output flag to print stdout of successful tests
This pull request adds a new flag `--show-output` for tests to show the output of successful tests. For most formatters this was already supported just not exposed via the CLI (apparently only used by `librustdoc`). I've also added support for this option in the JSON formatter.
This kind of fixes https://github.com/rust-lang/rust/issues/54669 which wants `--format json` to work with `--nocapture`, which is... well, impossible. What this issue really calls for is `--show-output` as implemented here.
This introduces a WithFormatter abstraction that permits one-time
fmt::Display on an arbitrary closure, created via `display_fn`. This
allows us to prevent allocation while still using functions instead of
structs, which are a bit unwieldy to thread arguments through as they
can't easily call each other (and are generally a bit opaque).
The eventual goal here is likely to move us off of the formatting
infrastructure entirely in favor of something more structured, but this
is a good step to move us in that direction as it makes, for example,
passing a context describing current state to the formatting impl much
easier.
Rollup of 6 pull requests
Successful merges:
- #63317 (Do not complain about unused code when used in `impl` `Self` type)
- #63693 (Fully implement or-pattern parsing)
- #63836 (VxWorks does not provide a way to set the task name except at creation time)
- #63845 (Removed a confusing FnOnce example)
- #63855 (Refactor feature gates)
- #63921 (add link to FileCheck docs)
Failed merges:
r? @ghost
