The channel's `Block::new` was causing a stack overflow because it held
32 item slots, instantiated on the stack before moving to `Box::new`.
The 32x multiplier made modestly-large item sizes untenable.
That block is now initialized directly on the heap.
Fixes#102246
Make fn_abi_sanity_check a bit stricter
The Rust ABI must ignore all ZST arguments, all ignored arguments must be either ZST or uninhabited. And finally ScalarPair should never be passed as PassMode::Direct.
Remove unused intercrate dependencies
Checked by enabling `-Wunused-crate-dependencies`
`driver_impl` still depends on `index` to forward the `rustc_randomized_layouts` feature, and `rustc_main` depends on several unused crates for sysroot reasons
r? compiler
Now that lints can add @eval_always at the end of their definition, the lint
declaration might not end right after the description. The `update_lints`
command can skip everything that comes after that.
The author lint is not an internal lint, and should also be enabled, when Clippy
is distributed through rustup. This moves the author lint test cases back to
tests/ui.
The Rust ABI must ignore all ZST arguments, all ignored arguments must
be either ZST or uninhabited. And finally ScalarPair should never be
passed as PassMode::Direct.
unpin and update memchr
I'm unable to build x86_64-pc-windows-gnu Rust due to some weird binutils bug, but thinlto issue seems to be no longer present. Let's give it a go on the CI.
Possibly fixed by https://github.com/rust-lang/rust/pull/129079Fixes#127890
chore(issue-template): fix branch name
It was renamed to "main" from "master".
It might seem not important (actually it is not important) but I think there is no need to see this every time when we clicked this link:
And this way is faster I guess.
fix(x): fix a regex used to find python executable
Isn't the regex `^python[2-3]\.[0-9]\+$` wrong? It doesn't match, for example, with `python2.8`. There should be a plus sign at the end for a match, like `python2.8+`. I think `[0-9]+` is meant here instead of `[0-9]\+`. In that case a string like `python2.8` would match. This wasn't noticed because the script probably find and run the Python executable before this line.
bootstrap: add quoting support to avoid splitting
With this change, it is now possible to pass quotes to the configure script, such as
`./configure.py --set=target.\"thumbv8m.main-none-eabi\".linker=/linker`
or
`./configure.py '--set=target."thumbv8m.main-none-eabi".linker=/linker'`
, which will treat `thumbv8.main-none-eabi` as a whole part. Currently, the string would be split into two elements: `thumbv8`, and `main-none-eabi`.
The approach taken is to perform custom splitting instead of using `str.split()` and then repairing the split. Also, There are numerous corner cases not handled: the custom split doesn't differentiate between single quotes or double quotes, so it is perfectly possible to pass `./configure.py --set=target.\"thumbv8m.main-none-eabi\'.linker=/linker` and the behaviour would be the same as with all double quotes or single quotes.
As for the code, i'm unsure on whether to delimit strings with double or single quotes. I've seen both single quotes and double quotes used to delimit strings, like in
```py
err("Option '{}' provided more than once".format(key))
```
and this a handful of lines down:
```py
if option.name == 'sccache':
set('llvm.ccache', 'sccache', config)
```
Please advise on the wanted one.
Fixes#130602
r? `@onur-ozkan`
Thanks in advance for the feedback!
Inline str::repeat
`str` is non-generic and `str.repeat()` doesn't get inlined, which makes it use a slower algorithm in case of 1-char repetitions. Equivalent byte slice does get inlined: https://rust.godbolt.org/z/4arvh97r4
Revert using `HEAP` static in Windows alloc
Fixes#131468
This does the minimum to remove the `HEAP` static that was causing chromium issues. It would be worth having a more substantial look at this module but for now I think this addresses the immediate issue.
cc `@danakj`
