Since commit 46068c9da, call to `reserve()` on empty vec allocates
exactly requested capacity, so unroll of first iteration may help only
with branch prediction.
Resolves #29672. This happened because rust runtime startup objects, rsbegin.o and rsend.o, were not included in the target libraries package for -windows-gnu.
r? @alexcrichton
The `enable-nonzeroing-move-hints` flag name was too long and caused misalignment of the help text.
Now calculating the needed padding dynamically from the available flags instead.
When merging two sorted blocks `left` and `right` if the last element in
`left` is <= the first in `right`, the blocks are already sorted.
Add this as an additional fast path by simply copying the whole left
block into the output and advancing the left pointer. The right block is
then treated the same way by the already present logic in the merge
loop.
Reduces runtime of .sort() to less than 50% of the previous, if the data
was already perfectly sorted. Sorted data with a few swaps are also
sorted quicker than before. The overhead of one comparison per merge
seems to be negligible.
Allow the changing of `target_family` through flexible configuration. The whole computing world isn't just a binary of *nix and Windows! Makes porting `libstd` and co to new platforms a lot less painful.
* Store the native representation directly in the `ExitStatus` structure instead
of a "parsed version" (mostly for Unix).
* On Windows, be more robust against processes exiting with the status of 259.
Unfortunately this exit code corresponds to `STILL_ACTIVE`, causing libstd to
think the process was still alive, causing an infinite loop. Instead the loop
is removed altogether and `WaitForSingleObject` is used to wait for the
process to exit.
* Store the native representation directly in the `ExitStatus` structure instead
of a "parsed version" (mostly for Unix).
* On Windows, be more robust against processes exiting with the status of 259.
Unfortunately this exit code corresponds to `STILL_ACTIVE`, causing libstd to
think the process was still alive, causing an infinite loop. Instead the loop
is removed altogether and `WaitForSingleObject` is used to wait for the
process to exit.
Handle them in `middle::reachable` instead (no optimizations so far, just drop all trait impl items into the reachable set, as before). Addresses the concerns from https://github.com/rust-lang/rust/pull/29291#discussion_r43672413
\+ In `middle::reachable` don't treat impls of `Drop` specially, they are subsumed by the general impl treatment.
\+ Add some tests checking reachability of trait methods written in UFCS form
\+ Minor refactoring in the second commit
r? @alexcrichton
Old doctest names
```bash
test sync::atomic::load_0 ... ok
test sync::atomic::load_0 ... ok
test sync::atomic::load_0 ... ok
test sync::atomic::load_0 ... ok
```
New doctest names
```bash
test sync::atomic::AtomicBool::load_0 ... ok
test sync::atomic::AtomicIsize::load_0 ... ok
test sync::atomic::AtomicPtr<T>::load_0 ... ok
test sync::atomic::AtomicUsize::load_0 ... ok
```
Introduce a `SwitchInt` and restructure pattern matching to collect integers and characters into one master switch. This is aimed at #29227, but is not a complete fix. Whereas before we generated an if-else-if chain and, at least on my machine, just failed to compile, we now spend ~9sec compiling `rustc_abuse`. AFAICT this is basically just due to a need for more micro-optimization of the matching process: perf shows a fair amount of time just spent iterating over the candidate list. Still, it seemed worth opening a PR with this step alone, since it's a big step forward.
Currently if a print happens while a thread is being torn down it may cause a
panic if the LOCAL_STDOUT TLS slot has been destroyed by that point. This adds a
guard to check and prints to the process stdout if that's the case (as we do for
if the slot is already borrowed).
Closes#29488
As discovered in #29298, `env::set_var("", "")` will panic, but it turns out
that it *also* deadlocks on Unix systems. This happens because if a panic
happens while holding the environment lock, we then go try to read
RUST_BACKTRACE, grabbing the environment lock, causing a deadlock.
Specifically, the changes made here are:
* The environment lock is pushed into `std::sys` instead of `std::env`. This
also only puts it in the Unix implementation, not Windows where the functions
are already threadsafe.
* The `std::sys` implementation now returns `io::Result` so panics are
explicitly at the `std::env` level.
