After #43252 is merged, building stage0 libcore with -i (--incremental)
flag will cause 17 "Quasi-quoting might make incremental compilation very
inefficient: NtExpr(..)" warnings, as in #40946.
Fixing the warning in #40946 will take 12 weeks from now to make into the
next stage0, so it is quicker to workaround it in libcore instead.
Add support for dylibs with Address Sanitizer
Many applications use address sanitizer to assert correct behaviour of their programs. When using Rust with C, it's much more important to assert correct programs with tools like asan/lsan due to the unsafe nature of the access across an ffi boundary. However, previously only rust bin types could use asan. This posed a challenge for existing C applications that link or dlopen .so when the C application is compiled with asan.
This PR enables asan to be linked to the dylib and cdylib crate type. We alter the test to check the proc-macro crate does not work with -Z sanitizer=address. Finally, we add a test that compiles a shared object in rust, then another rust program links it and demonstrates a crash through the call to the library.
This PR is nearly complete, but I do require advice on the change to fix the -lasan that currently exists in the dylib test. This is required because the link statement is not being added correctly to the rustc build when -Z sanitizer=address is added (and I'm not 100% sure why)
Thanks,
Now there's a way to add suggestions that hide the suggested code when
presented inline, to avoid weird wording when short code snippets are
added at the end.
The produced paths aren't stable between builds, since
reporting paths inside resolve, before resolve is finished
might produce paths resolved to type aliases instead of
the concrete type.
Compile-fail tests can match just parts of messages, so they
don't "suffer" from this issue.
This is just a workaround, the instability should be fixed
in the future.
Stabilize float_bits_conv for Rust 1.21
Stabilizes the `float_bits_conv` lib feature for the 1.20 release of Rust. I've initially implemented the feature in #39271 and later made PR #43025 to output quiet NaNs even on platforms with different encodings, which seems to have been the only unresolved issue of the API.
Due to PR #43025 being only applied to master this stabilisation can't happen for Rust 1.19 through the usual "stabilisation on beta" system that is being done for library APIs.
r? @BurntSushi
closes#40470.
When triggering type ascription in such a way that we can infer a
statement end was intended, add a suggestion for the change. Always
point out the reason for the expectation of a type is due to type
ascription.
Document default values for primitive types
All primitive types implement the `Default` trait but the documentation just says `Returns the "default value" for a type.` and doesn't give a hint about the actual default value. I think it would be good to document the default values in a proper way.
I changed the `default_impl` macro to accept a doc string as a third parameter and use this string to overwrite the documentation of `default()` for each primitive type.
The generated documentation now looks like this:
We want the suggested replacement (which IDE tooling and such might offer to
automatically swap in) to, like, actually be correct: suggesting `MyVariant(x)`
when the actual fix is `MyEnum::MyVariant(x)` might be better than nothing, but
Rust is supposed to be the future of computing: we're better than better than
nothing.
As an exceptional case, we excise the prelude path, preferring to suggest
`Some` or `Ok` rather than `std::prelude::v1::Some` and
`std::prelude::v2::Ok`. (It's not worth the effort to future-proof against
hypothetical preludes v2, v3, &c.: we trust our successors to grep—excuse me,
ripgrep—for that.)
Also, don't make this preëmpt the existing probe-for-return-type suggestions,
despite their being looked unfavorably upon, at least in this situation
(https://github.com/rust-lang/rust/issues/42764#issuecomment-311388958): Cody
Schafer pointed out that that's a separate issue
(https://github.com/rust-lang/rust/pull/43178#issuecomment-314953229).
This is in the matter of #42764.
Update the `cargo` submodule
Notably pull in an update to the `jobserver` crate to have Cargo set the
`CARGO_MAKEFLAGS` environment variable instead of the `MAKEFLAGS` environment
variable.
cc https://github.com/rust-lang/rust/issues/42635
support pub(restricted) in thread_local! (round 2)
Resurrected #40984 now that the issue blocking it was fixed. Original description:
`pub(restricted)` was stabilized in #40556 so let's go!
Here is a [playground](https://play.rust-lang.org/?gist=f55f32f164a6ed18c219fec8f8293b98&version=nightly&backtrace=1).
I changed the interface of `__thread_local_inner!`, which is supposedly unstable but this is not checked for macros (#34097 cc @petrochenkov @jseyfried), so this may be an issue.
windows::fs::symlink_dir: fix example to actually use symlink_dir
I don't have a windows machine, so I couldn't test if this doctest still works -- but it looks trivial enough. (I know, famous last words.)
Disable big-endian simd in swap_nonoverlapping_bytes
This is a workaround for #42778, which was git-bisected to #40454's
optimizations to `mem::swap`, later moved to `ptr` in #42819. Natively
compiled rustc couldn't even compile stage1 libcore on powerpc64 and
s390x, but they work fine without this `repr(simd)`. Since powerpc64le
works OK, it seems probably related to being big-endian.
The underlying problem is not yet known, but this at least makes those
architectures functional again in the meantime.
cc @arielb1
Forward more Iterator methods
This allows in more cases to take advantage of specific (possibly more optimized) impls of these methods, rather than the default one defined for all `Iterator`s.
I also wanted to do this for `&mut I` and `Box<I>`, but that didn’t compile for two reasons:
* To make the trait object-safe, generic methods (e.g. that take a closure parameter) have a `where Self: Sized` bound. But e.g. `Box<I>: Sized` does not imply `I: Sized`, and adding an additional bound in the impl is not allowed. Some for of specialization would be needed here.
* With e.g. a `F: FnMut(Self::Item) -> bool` bound and a `type Item = I::Item` associated types, I got errors like `F does not implement FnMut(I::Item) -> bool`. This looks like a limitation in the trait resolution system not recognizing that `Self::Item == I::Item` or "propagating" that fact to `FnMut` bounds.
This allows changes to the Rust language that have both library
and language components share one feature gate.
The feature gates need to be "about the same change", so that both
library and language components must either be both unstable, or
both stable, and share the tracking issue.
Removes the ugly "proc_macro" exception.
Closes#43089
integrate anon dep nodes into trait selection
Use anonymous nodes for trait selection. In all cases, we use the same basic "memoization" strategy:
- Store the `DepNodeIndex` in the slot along with value.
- If value is present, return it, and add a read of the dep-node-index.
- Else, start an anonymous task, and store resulting node.
We apply this strategy to a number of caches in trait selection:
- The "trans" caches of selection and projection
- The "evaluation" cache
- The "candidate selection" cache
In general, for our cache strategy to be "dep-correct", the computation of the value is permitted to rely on the *value in the key* but nothing else. The basic argument is this: in order to look something up, you have to produce the key, and to do that you must have whatever reads were needed to create the key. Then, you get whatever reads were further needed to produce the value. But if the "closure" that produced the value made use of *other* environmental data, not derivable from the key, that would be bad -- but that would **also** suggest that the cache is messed up (though it's not proof).
The structure of these caches do not obviously prove that the correctness criteria are met, and I aim to address that in further refactorings. But I *believe* it to be the case that, if we assume that the existing caches are correct, there are also no dependency failures (in other words, if there's a bug, it's a pre-existing one). Specifically:
- The trans caches: these take as input just a `tcx`, which is "by definition" not leaky, the `trait-ref` etc, which is part of the key, and sometimes a span (doesn't influence the result). So they seem fine.
- The evaluation cache:
- This computation takes as input the "stack" and has access to the infcx.
- The infcx is a problem -- would be better to take the tcx -- and this is exactly one of the things I plan to improve in later PRs. Let's ignore it for now. =)
- The stack itself is also not great, in that the *key* only consists of the top-entry in the stack.
- However, the stack can cause a problem in two ways:
- overflow (we panic)
- cycle check fails (we do not update the cache, I believe)
- The candidate selection cache:
- as before, takes the "stack" and has access to the infcx.
- here it is not as obvious that we avoid caching stack-dependent computations. However, to the extent that we do, this is a pre-existing bug, in that we are making cache entries we shouldn't.
- I aim to resolve this by -- following the chalk-style of evaluation -- merging candidate selection and evaluation.
- The infcx is a problem -- would be better to take the tcx -- and this is exactly one of the things I plan to improve in later PRs. Let's ignore it for now. =)
- The stack itself is also not great, in that the *key* only consists of the top-entry in the stack.
- Moreover, the stack would generally just introduce ambiguities and errors anyhow, so that lessens the risk.
Anyway, the existing approach to handle dependencies in the trait code carries the same risks or worse, so this seems like a strict improvement!
r? @michaelwoerister
cc @arielb1
