travis: Ensure cargo links libcurl statically
We don't want a dynamic dependency in the library that we ship, so link it
statically by configuring curl-sys's build script to not pick up the system
version via pkg-config.
Ignore the type of error altogether. The rationale is: it doesn't matter
what was the problem if the directory is there. In the previous versions
if the directory was there already we wouldn't even attempt to create
it, so we wouldn't know about the problem neither.
Make test path length smaller in `concurrent_recursive_mkdir` test.
It is more robust to not fail if any directory in a path was created
concurrently. This change lifts rustc internal `create_dir_racy` that
was created to handle such conditions to be new `create_dir_all`
implementation.
Drop of arrays is now translated in trans::block in an ugly way that I
should clean up in a later PR, and does not handle panics in the middle
of an array drop, but this commit & PR are growing too big.
These changes are in the same commit to avoid needing to adapt
meth::trans_object_shim to the new scheme.
One codegen-units test is broken because we instantiate the shims even
when they are not needed. This will be fixed in the next PR.
rustbuild: Retry downloads of OpenSSL source
We need this to compile Cargo and we download it at build time, but as like all
other network requests it has a chance of failing. This commit moves the source
of the tarball to a mirror (S3 seems semi-more-reliable most of the time) and
also wraps the download in a retry loop.
cc #40474
Inline functions Ordering::{then, then_with}
@jongiddy noticed bad performance due to the lack of inlining on `then`
and `then_with`. I confirmed that inlining really is the culprit by
creating a custom `then` function and repeating his benchmark on my
machine with and without the `#[inline]` attribute.
The numbers were exactly the same on my machine without the attribute.
With `#[inline]` I got the same performance as I did with manually
inlined implementation.
The problem was reported in #37053.
A few improvements to the `core::hash` top-level docs.
Primarily opened to address the concerns brought up in
https://github.com/rust-lang/rust/issues/40498.
* run rustfmt on code blocks
* use `DefaultHasher` instead of deprecated `SipHasher`
* rename `hash` to `calculate_hash` to prevent confusion with the `hash`
method
std: remove a workaround for privacy limitations
`std:🧵:Thread` implements a non-exported `NewThread` trait to allow for internal-only use of `Thread::new`. Nowadays we have `pub(crate)`, which accomplishes the same thing but much more idiomatically.
Rustdoc handles this correctly (I checked and I didn't see `Thread::new` on the rustdoc entry for `Thread` with this change), and the stage1 `rustc` emits the correct error still (I'm assuming that the stage1 compiler uses my `libstd`?):
```
$ ./build/x86_64-apple-darwin/stage1/bin/rustc test.rs
error: method `new` is private
--> test.rs:4:18
|
4 | let thread = thread::Thread::new(None);
| ^^^^^^^^^^^^^^^^^^^
error: aborting due to previous error
```
