Remove dep-info files as targets in themselves
If you ask `rustc` to `--emit dep-info`, the resulting dependency file contains a rule for producing the dependency file itself. This differs from the output of `gcc -MD` or `clang -MD`, which only includes dependency rules for the object files produced.
Tools like Ninja often consume and delete dependency files as soon as they’re produced for performance reasons, particularly on Windows. In the case of `rustc` output, though, the recently-deleted dependency file is cached by Ninja as a target, and therefore triggers a rebuild every time.
This very small patch removes the dep-info file from the list of output filenames, so it matches the behavior of gcc and clang.
Pull in rust-lang/rust-installer#76 to get streamed tarball generation,
rather than batching it all in memory, while still getting the benefit
of compressing in parallel.
This fixes an accidental regression #46335 where the behavior of
`Path::ends_with` is different from `str::ends_with` (paths operate over
components, strs operate over chars).
As discussed in #47427, let's not have a separate container for doing
CloudABI builds. It's a lot faster if we integrate it into an existing
container, so there's less duplication of what's being built.
Upgrade the existing container to Ubuntu 17.10, which is required for
CloudABI builds. The version of Clang shipped with 16.04 is not recent
enough to support CloudABI properly.
The match lowering code, when lowering matches against bytestrings,
works by coercing both the scrutinee and the pattern to `&[u8]` and
then comparing them using `<[u8] as Eq>::eq`.
If the scrutinee is already of type `&[u8]`, then unsizing it is both
unneccessary and a trait error caught by the new and updated MIR typeck,
so this PR changes lowering to avoid doing that (match lowering tried to
avoid that before, but that attempt was quite broken).
Fixes#46920.
Setting up a cross compilation toolchain for CloudABI is relatively
easy. It's just a matter of installing a somewhat recent version of
Clang (5.0 preferred) and installing the corresponding
${target}-cxx-runtime package, containing a set of core C/C++ libraries
(libc, libc++, libunwind, etc).
Eventually it would be nice if we could also run 'x.py test'. That,
however still requires some more work. Both libtest and compiletest
would need to be adjusted to deal with CloudABI's requirement of having
all of an application's dependencies injected. Let's settle for just
doing 'x.py dist' for now.
The function parse_cfg_prefix() is not really parsing. It's just checking
whether the prefix is present or not. So the new function name as suggested by
@Mark-Simulacrum is better.
Previous patch introduced something like if x {true} else {false} which can be
simply replaced by returning x here.
Thanks to @kennytm for spotting it.
This patch implements "only-<platforms>" for tests headers using which one can
specify just the platforms on which the test should run rather than listing all
the platforms to ignore using "ignore-<platforms>".
This is a fix for issues #33581 and #47459.
The incompetent fool who added these suggestions in 38e5a964f2 apparently
thought it was safe to assume that, because the offending function or
static was unreachable, it would therefore have not have any existing
visibility modifiers, making it safe for us to unconditionally suggest
inserting `pub`. This isn't true.
This resolves#47383.
rustc_trans: reorganize CrateContext and rename context types.
Firstly, the `{Shared,Local}CrateContext` hasn't been meaningful for a while now, and this PR resolves it by moving all their fields to `CrateContext` and removing redundant accessor methods.
Secondly, this PR contains the following mass-renames:
* `ccx: CrateContext` -> `cx: CodegenCx`
* `mircx: MirContext` -> `fx: FunctionCx`
* `bcx: Builder` -> `bx: Builder`
r? @nikomatsakis
Avoid overlapping spans by only pointing at the arguments that are not
being used in the argument string. Enable libsyntax to have diagnostics
with multiple primary spans by accepting `Into<MultiSpan>` instead of
`Span`.
These are already removed in the normal optimization pipeline - so this
should slightly improve codegen performance, as these cleanup blocks are
known to hurt LLVM.
This un-regresses and is therefore a fix for #47442. However, the
reporter of that issue should try using `-C panic=abort` instead of
carefully avoiding panics.
To workaround a bug in glibc <= 2.26 lookup_host() calls res_init()
based on the glibc version detected at runtime. While this avoids
calling res_init() on platforms where it's not required we will still
end up linking against the symbol.
This causes an issue on macOS where res_init() is implemented in a
separate library (libresolv.9.dylib) from the main libc. While this is
harmless for standalone programs it becomes a problem if Rust code is
statically linked against another program. If the linked program doesn't
already specify -lresolv it will cause the link to fail. This is
captured in issue #46797
Fix this by hooking in to the glibc workaround in `cvt_gai` and only
activating it for the "gnu" environment on Unix This should include all
glibc platforms while excluding musl, windows-gnu, macOS, FreeBSD, etc.
This has the side benefit of removing the #[cfg] in sys_common; only
unix.rs has code related to the workaround now.
