The makefiles and the windows installer disagree on the name of this file. In practical terms this change only means that the '-pre' installers will be named 'rust-0.9-pre-install.exe' instead 'rust-0.9-install.exe'.
libnative erroneously would attempt to fill the entire buffer in a call to
`read` before returning, when rather it should return immediately because
there's not guaranteed to be any data that will ever be received again.
Close#11328
Rearrange some code in rustpkg:
* Move command line argument parsing to parse_args.rs
* Introduce an enum to descibe commands such as Build, Install, and Clean.
* Move sysroot from Context to BuildContext, to make parse_args more modular.
This is my first pull request, so please tell me if there is anything I need to do.
The primary user of `@T`/`P<T>` references from `Visitor` was `ast_map`, which in turn had two users (phase 3 in rustc and a step in loading items from metadata).
Both of them have been rewritten to use `ast_map` as a folder (this might speed up the compilation time of stage2 rustc by 100-200ms just because a fold + a visit are merged together).
This requires a single `*mut` pointer to implement; I've justified its existence & correctness in the code.
Also, converts the mutable and immutable iterators to share code with some macro ~~madness~~ manipulation.
There was a scheduling race where a child may not increment the global task
count before the parent exits, and the parent would then think that there are no
more tasks left.
Closes#11039
libnative erroneously would attempt to fill the entire buffer in a call to
`read` before returning, when rather it should return immediately because
there's not guaranteed to be any data that will ever be received again.
Close#11328
There was a scheduling race where a child may not increment the global task
count before the parent exits, and the parent would then think that there are no
more tasks left.
This reverts commit f1b5f59287.
Using a private function of a library is a bad idea: several people (on
Linux) were meeting with linking errors because of it (different/older
versions of glibc).
If the main closure failed, then the `exit_code` variable would still be `None`,
and the `unwrap()` was failing (triggering a process abort). This changes the
`unwrap()` to an `unwrap_or()` in order to prevent process abort and detect when
the native task failed.
This is causing maybe_get_doc to fail when called at the top level of
the metadata and doesn't appear to be needed anymore.
A try build succeeded so this should be okay.
- don't check for an hardcoded copyright claim year, check the 2 surrounding strings instead
- logic: if either the `//` or `#`-style copyright patterns are found, don't invalidate
- cleanup hardcoded content and streamline the few files with different line breaks
r? @brson
This removes the feature where newtype structs can be dereferenced like pointers, and likewise where certain enums can be dereferenced (which I imagine nobody realized still existed). This ad-hoc behavior is to be replaced by a more general overloadable dereference trait in the future.
I've been nursing this patch for two months and think it's about rebased up to master.
@nikomatsakis this makes a bunch of your type checking code noticeably uglier.
If there is a lot of data in thread-local storage some implementations
of pthreads (e.g. glibc) fail if you don't request a stack large enough
-- by adjusting for the minimum size we guarantee that our stacks are
always large enough. Issue #6233.
