This has a number of advantages compared to creating a copy in memory
and passing a pointer. The obvious one is that we don't have to put the
data into memory but can keep it in registers. Since we're currently
passing a pointer anyway (instead of using e.g. a known offset on the
stack, which is what the `byval` attribute would achieve), we only use a
single additional register for each fat pointer, but save at least two
pointers worth of stack in exchange (sometimes more because more than
one copy gets eliminated). On archs that pass arguments on the stack, we
save a pointer worth of stack even without considering the omitted
copies.
Additionally, LLVM can optimize the code a lot better, to a large degree
due to the fact that lots of copies are gone or can be optimized away.
Additionally, we can now emit attributes like nonnull on the data and/or
vtable pointers contained in the fat pointer, potentially allowing for
even more optimizations.
This results in LLVM passes being about 3-7% faster (depending on the
crate), and the resulting code is also a few percent smaller, for
example:
|text|data|filename|
|----|----|--------|
|5671479|3941461|before/librustc-d8ace771.so|
|5447663|3905745|after/librustc-d8ace771.so|
| | | |
|1944425|2394024|before/libstd-d8ace771.so|
|1896769|2387610|after/libstd-d8ace771.so|
I had to remove a call in the backtrace-debuginfo test, because LLVM can
now merge the tails of some blocks when optimizations are turned on,
which can't correctly preserve line info.
Fixes#22924
Cc #22891 (at least for fat pointers the code is good now)
This has a number of advantages compared to creating a copy in memory
and passing a pointer. The obvious one is that we don't have to put the
data into memory but can keep it in registers. Since we're currently
passing a pointer anyway (instead of using e.g. a known offset on the
stack, which is what the `byval` attribute would achieve), we only use a
single additional register for each fat pointer, but save at least two
pointers worth of stack in exchange (sometimes more because more than
one copy gets eliminated). On archs that pass arguments on the stack, we
save a pointer worth of stack even without considering the omitted
copies.
Additionally, LLVM can optimize the code a lot better, to a large degree
due to the fact that lots of copies are gone or can be optimized away.
Additionally, we can now emit attributes like nonnull on the data and/or
vtable pointers contained in the fat pointer, potentially allowing for
even more optimizations.
This results in LLVM passes being about 3-7% faster (depending on the
crate), and the resulting code is also a few percent smaller, for
example:
text data filename
5671479 3941461 before/librustc-d8ace771.so
5447663 3905745 after/librustc-d8ace771.so
1944425 2394024 before/libstd-d8ace771.so
1896769 2387610 after/libstd-d8ace771.so
I had to remove a call in the backtrace-debuginfo test, because LLVM can
now merge the tails of some blocks when optimizations are turned on,
which can't correctly preserve line info.
Fixes#22924
Cc #22891 (at least for fat pointers the code is good now)
As per #26009 this PR implements a new collation system for extended-error metadata. I've tried to keep it as simple as possible for now, so there's no uniqueness checking and minimal modularity.
Although using a lint was discussed in #26009 I decided against this because it would require converting the AST output from the plugin back into an internal data-structure. Emitting the metadata from within the plugin prevents this double-handling. We also didn't identify this as the source of the failures last time, although something untoward was definitely happening... With that in mind I would like as much feedback as possible on this before it's merged, I don't want to break the bots again!
I've successfully built for my host architecture and I'm building an ARM cross-compiler now to test my assumptions about the various `CFG` variables. Despite the confusing name of `CFG_COMPILER_HOST_TRIPLE` it is actually the compile time target triple, as explained in `mk/target.mk`.
```
# This is the compile-time target-triple for the compiler. For the compiler at
# runtime, this should be considered the host-triple. More explanation for why
# this exists can be found on issue #2400
export CFG_COMPILER_HOST_TRIPLE
```
CC @pnkfelix @brson @nrc @alexcrichton
Closes#25705, closes#26009.
It now says '#[feature] may not be used on the stable release channel'.
I had to convert this error from a lint to a normal compiler error.
I left the lint previously-used for this in place since removing it is
a breaking change. It will just go unused until the end of time.
Fixes#24125
Environment variables are global state so this can lead to surprising results if
the driver is called in a multithreaded environment (e.g. doctests). There
shouldn't be any memory corruption that's possible, but a lot of the bots have
been failing because they can't find `cc` or `gcc` in the path during doctests,
and I highly suspect that it is due to the compiler modifying `PATH` in a
multithreaded fashion.
This commit moves the logic for appending to `PATH` to only affect the child
process instead of also affecting the parent, at least for the linking stage.
When loading dynamic libraries the compiler still modifies `PATH` on Windows,
but this may be more difficult to fix than spawning off a new process.
The execv family of functions do not modify their arguments, so they do
not need mutable pointers. The C prototypes take a constant array of
mutable C-strings, but that's a legacy quirk from before C had const
(since C string literals have type `char *`). The Rust prototypes had
`*mut` in the wrong place, anyway: to match the C prototypes, it should
have been `*const *mut c_char`. But it is safe to pass constant strings
(like string literals) to these functions.
getopt is a special case, since GNU getopt modifies its arguments
despite the `const` claim in the prototype. It is apparently only
well-defined to call getopt on the actual argc and argv parameters
passed to main, anyway. Change it to take `*mut *mut c_char` for an
attempt at safety, but probably nobody should be using it from Rust,
since there's no great way to get at the parameters as passed to main.
Also fix the one caller of execvp in libstd, which now no longer needs
an unsafe cast.
Fixes#16290.
[Randard](https://github.com/brson/rust/blob/relnotes/RELEASES.md).
I need help identifying language things and highlights, and I am kind of tired of the `~X changes, numerous bugfixes` thing, which was cute for a few years.
cc @aturon @nrc @pnkfelix
Expand the "givens" set to cover transitive relations. The givens array
stores relationships like `'c <= '0` (where `'c` is a free region and
`'0` is an inference variable) that are derived from closure
arguments. These are (rather hackily) ignored for purposes of inference,
preventing spurious errors. The current code did not handle transitive
cases like `'c <= '0` and `'0 <= '1`. Fixes#24085.
r? @pnkfelix
cc @bkoropoff
*But* I am not sure whether this fix will have a compile-time hit. I'd like to push to try branch observe cycle times.
Pre-requisite for splitting the type context into global and local parts.
The `Repr` and `UserString` traits were also replaced by `Debug` and `Display`.
stores relationships like `'c <= '0` (where `'c` is a free region and
`'0` is an inference variable) that are derived from closure
arguments. These are (rather hackily) ignored for purposes of inference,
preventing spurious errors. The current code did not handle transitive
cases like `'c <= '0` and `'0 <= '1`. Fixes#24085.
