spawn() is expected to return an error if the specified file could not be
executed. FreeBSD's posix_spawn() supports returning ENOENT/ENOEXEC if
the exec() fails, which not all platforms support. This brings a very
significant performance improvement for FreeBSD, involving heavy use of
Command in threads, due to fork() invoking jemalloc fork handlers and
causing lock contention. FreeBSD's posix_spawn() avoids this problem
due to using vfork() internally.
This is the ideal FileType on Windows. You may not like it, but this is what peak performance looks like.
Theoretically this would fix https://github.com/rust-lang/rust/issues/46484
The current iteration of this PR should not cause existing code to break, but instead merely improves handling around reparse points. Specifically...
* Reparse points are considered to be symbolic links if they have the name surrogate bit set. Name surrogates are reparse points that effectively act like symbolic links, redirecting you to a different directory/file. By checking for this bit instead of specific tags, we become much more general in our handling of reparse points, including those added by third parties.
* If something is a reparse point but does not have the name surrogate bit set, then we ignore the fact that it is a reparse point because it is actually a file or directory directly there, despite having additional handling by drivers due to the reparse point.
* For everything which is not a symbolic link (including non-surrogate reparse points) we report whether it is a directory or a file based on the presence of the directory attribute bit.
* Notably this still preserves invariant that when `is_symlink` returns `true`, both `is_dir` and `is_file` will return `false`. The potential for breakage was far too high.
* Adds an unstable `FileTypeExt` to allow users to determine whether a symbolic link is a directory or a file, since `FileType` by design is incapable of reporting this information.
Use a range to identify SIGSEGV in stack guards
Previously, the `guard::init()` and `guard::current()` functions were
returning a `usize` address representing the top of the stack guard,
respectively for the main thread and for spawned threads. The `SIGSEGV`
handler on `unix` targets checked if a fault was within one page below that
address, if so reporting it as a stack overflow.
Now `unix` targets report a `Range<usize>` representing the guard memory,
so it can cover arbitrary guard sizes. Non-`unix` targets which always
return `None` for guards now do so with `Option<!>`, so they don't pay any
overhead.
For `linux-gnu` in particular, the previous guard upper-bound was
`stackaddr + guardsize`, as the protected memory was *inside* the stack.
This was a glibc bug, and starting from 2.27 they are moving the guard
*past* the end of the stack. However, there's no simple way for us to know
where the guard page actually lies, so now we declare it as the whole range
of `stackaddr ± guardsize`, and any fault therein will be called a stack
overflow. This fixes#47863.
Previously, the `guard::init()` and `guard::current()` functions were
returning a `usize` address representing the top of the stack guard,
respectively for the main thread and for spawned threads. The `SIGSEGV`
handler on `unix` targets checked if a fault was within one page below
that address, if so reporting it as a stack overflow.
Now `unix` targets report a `Range<usize>` representing the guard
memory, so it can cover arbitrary guard sizes. Non-`unix` targets which
always return `None` for guards now do so with `Option<!>`, so they
don't pay any overhead.
For `linux-gnu` in particular, the previous guard upper-bound was
`stackaddr + guardsize`, as the protected memory was *inside* the stack.
This was a glibc bug, and starting from 2.27 they are moving the guard
*past* the end of the stack. However, there's no simple way for us to
know where the guard page actually lies, so now we declare it as the
whole range of `stackaddr ± guardsize`, and any fault therein will be
called a stack overflow. This fixes#47863.
Only link res_init() on GNU/*nix
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.
Before this commit:
```shell
> cat main.rs
use std::net::ToSocketAddrs;
#[no_mangle]
pub extern "C" fn resolve_test() -> () {
let addr_list = ("google.com.au", 0).to_socket_addrs().unwrap();
println!("{:?}", addr_list);
}
> rustc --crate-type=staticlib main.rs
> clang libmain.a test.c -o combined
Undefined symbols for architecture x86_64:
"_res_9_init", referenced from:
std::net::lookup_host::h93c17fe9ad38464a in libmain.a(std-826c8d3b356e180c.std0.rcgu.o)
ld: symbol(s) not found for architecture x86_64
clang-5.0: error: linker command failed with exit code 1 (use -v to see invocation)
```
Afterwards:
```shell
> rustc --crate-type=staticlib main.rs
> clang libmain.a test.c -o combined
> ./combined
IntoIter([V4(172.217.25.131:0)])
```
Fixes #46797
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.
As discussed in #47268, libstd isn't ready to have certain functionality
disabled yet. Follow wasm's approach of adding no-op modules for all of
the features that we can't implement.
I've placed all of those shims in a shims/ subdirectory, so we (the
CloudABI folks) can experiment with removing them more easily. It also
ensures that the code that does work doesn't get polluted with lots of
useless boilerplate code.
Though CloudABI is strongly inspired by POSIX, its absence of features
that don't work well with capability-based sandboxing makes it different
enough that adding bits to sys/unix will make things a mess. This change
therefore adds CloudABI specific platform code under sys/cloudabi and
borrows parts from sys/unix that can be used without changes.
One of the goals of this implementation is to build as much as possible
directly on top of CloudABI's system call layer, as opposed to using the
C library. This is preferred, as the system call layer is supposed to be
stable, whereas the C library ABI technically is not. An advantage of
this approach is that it allows us to implement certain interfaces, such
as mutexes and condition variables more optimally. They can be lighter
than the ones provided by pthreads.
This change disables some modules that cannot realistically be
implemented right now. For example, libstd's pathname abstraction is not
designed with POSIX *at() (e.g., openat()) in mind. The *at() functions
are the only set of file system APIs available on CloudABI. There is no
global file system namespace, nor a process working directory.
Discussions on how to port these modules over are outside the scope of
this change.
Apart from this change, there are still some other minor fixups that
need to be made to platform independent code to make things build. These
will be sent out separately, so they can be reviewed more thoroughly.
These automatically generated Rust source files allow us to invoke
system calls within CloudABI processes. These will be used by libstd to
implement primitives for I/O, threading, etc.
These source files are normally part of the 'cloudabi' crate. In the
case of libstd, we'd better copy them into the source tree, as having
external dependencies in libstd is a bit messy. Original source files
can be found here:
https://github.com/NuxiNL/cloudabi/tree/master/rust
Replace empty array hack with repr(align)
As a side effect, this fixes the warning about repr(C, simd) that has been reported during x86_64 windows builds since #47111 (see also: #47103)
r? @alexcrichton
Redox - Implement rename using new system call
This does the following:
- Update syscall module to match upstream
- Implement rename using new system call
- Make readlink and symlink utilize O_CLOEXEC
- Make readlink and symlink not leave dangling file handles on failure
Convert warning about `*const _` to a future-compat lint
#46664 was merged before I could convert the soft warning about method lookup on `*const _` into a future-compatibility lint. This PR makes that change.
fixes#46837
tracking issue for the future-compatibility lint: #46906
r? @arielb1
