The must-use lint needs the DefId of called functions and method
receivers in order to look for a `#[must_use]` attribute, but this would
ICE (!) if a called function was actually a closure (with a non-unit
return value). Instead, let's be specific that we want a `Def::Fn`,
rather than blithely assuming that we can get the DefId of a qpath.
Supporting must-use closures doesn't seem like a priority, but could
conceivably be added in the future if desired (conditional on the
statement and expression attributes (#15701) story being amicable).
Fix ICE #48984
* ~~fbf6423 The tail type was not normalized.~~
* d0839d5680 The method had a wrong assumption that something whose parent is a trait is an associated item. Fixes#48984.
Add `-C target-feature` to all functions
Previously the features specified to LLVM via `-C target-feature` were only
reflected in the `TargetMachine` but this change *also* reflects these and the
base features inside each function itself. This change matches clang and...
Closesrust-lang-nursery/stdsimd#427
Previously the features specified to LLVM via `-C target-feature` were only
reflected in the `TargetMachine` but this change *also* reflects these and the
base features inside each function itself. This change matches clang and...
Closesrust-lang-nursery/stdsimd#427
Implement LazyBTreeMap and use it in a few places.
This is a thin wrapper around BTreeMap that avoids allocating upon creation.
I would prefer to change BTreeMap directly to make it lazy (like I did with HashSet in #36734) and I initially attempted that by making BTreeMap::root an Option<>. But then I also had to change Iter and Range to handle trees with no root, and those types have stability markers on them and I wasn't sure if that was acceptable. Also, BTreeMap has a lot of complex code and changing it all was challenging, and I didn't have high confidence about my general approach.
So I prototyped this wrapper instead and used it in the hottest locations to get some measurements about the effect. The measurements are pretty good!
- Doing a debug build of serde, it reduces the total number of heap allocations from 17,728,709 to 13,359,384, a 25% reduction. The number of bytes allocated drops from 7,474,672,966 to 5,482,308,388, a 27% reduction.
- It gives speedups of up to 3.6% on some rustc-perf benchmark jobs. crates.io, futures, and serde benefit most.
```
futures-check
avg: -1.9% min: -3.6% max: -0.5%
serde-check
avg: -2.1% min: -3.5% max: -0.7%
crates.io-check
avg: -1.7% min: -3.5% max: -0.3%
serde
avg: -2.0% min: -3.0% max: -0.9%
serde-opt
avg: -1.8% min: -2.9% max: -0.3%
futures
avg: -1.5% min: -2.8% max: -0.4%
tokio-webpush-simple-check
avg: -1.1% min: -2.2% max: -0.1%
futures-opt
avg: -1.2% min: -2.1% max: -0.4%
piston-image-check
avg: -0.8% min: -1.1% max: -0.3%
crates.io
avg: -0.6% min: -1.0% max: -0.3%
```
@Gankro, how do you think I should proceed here? Is leaving this as a wrapper reasonable? Or should I try to make BTreeMap itself lazy? If so, can I change the representation of Iter and Range?
Thanks!
rustc_driver: Catch ICEs on the main thread too
#48575 introduced an optimization to run rustc directly on the main thread when possible. However, the threaded code detects panics when they `join()` to report as an ICE. When running directly, we need to use `panic::catch_unwind` to get the same effect.
cc @ishitatsuyuki
r? @alexcrichton
Added warning for unused arithmetic expressions
The compiler now displays a warning when a binary arithmetic operation is evaluated but not used. This resolves#50124 by following the instructions outlined in the issue. The changes are as follows:
- Added new pattern matching for unused arithmetic expressions in `src/librustc_lint/unused.rs`
- Added `#[must_use]` attributes to the binary operation methods in `src/libcore/internal_macros.rs`
- Added `#[must_use]` attributes to the non-assigning binary operators in `src/libcore/ops/arith.rs`
Access individual fields of tuples, closures and generators on drop.
Fixes#48623, by extending the change in #47917 to closures. Also does this for tuples and generators for consistency.
Enums are unchanged because there is now way to borrow `*enum.field` without borrowing `enum.field` at the moment, so any error would be reported when the enum goes out of scope. Unions aren't changed because unions they don't automatically drop their fields.
r? @nikomatsakis
rustc: Emit `uwtable` for allocator shims
This commit emits the `uwtable` attribute to LLVM for platforms that require it
for the allocator shims that we generate to ensure that they can hopefully get
unwound past. This is a stab in the dark at helping
https://bugzilla.mozilla.org/show_bug.cgi?id=1456150 along.
rustc: Disable threads in LLD for wasm
Upstream bug reports (rustwasm/wasm-bindgen#119) show that this may be the
culprit of odd crashes/hangs. The linker is a tiny fraction of build time anyway
right now so let's disable it and figure out how to possibly reenable it later
if necessary.
Add some utilities to `libsyntax`
Adds a few functions to `Mark` and `Span` that I found useful in an upcoming refactor of NLL region error reporting. Also includes some new documentation based on my discussion with @jseyfried on IRC.
r? @jseyfried
Remove hack around comparisons of i1 values (fixes#40980)
The regression test still passes without that 2 years old hack. The underlying
LLVM bug has probably been fixed upstream since then.
