Don't ICE on item in `.await` expression
The code for lowering a `.await` expression missed that item IDs may already have been assigned for items inside of an `async` block, or for closures. This change means we no longer exit early after finding a `.await` in a block that isn't `async` and instead just emit the error. This avoids an ICE generated due to item IDs not being densely generated. (The `YieldSource` of the generated `yield` expression is used to avoid errors generated about having `yield` expressions outside of generator literals.)
r? @cramertj
Resolves#62009 and resolves#61685
Add suggestion for missing `.await` keyword
This commit adds a suggestion diagnostic for missing `.await`. In order to do this, the trait `Future` is promoted to a lang item.
Compiling code of the form:
```rust
#![feature(async_await)]
fn take_u32(x: u32) {}
async fn make_u32() -> u32 {
22
}
async fn foo() {
let x = make_u32();
take_u32(x)
}
fn main() {}
```
Will now result in the suggestion:
```
error[E0308]: mismatched types
--> src/main.rs:11:18
|
11 | take_u32(x)
| ^
| |
| expected u32, found opaque type
| help: consider using `.await` here: `x.await`
|
= note: expected type `u32`
found type `impl std::future::Future`
```
This commit does not cover chained expressions and therefore only covers the case originally pointed out in #61076. Cases I can think of that still need to be covered:
- [ ] Return places for functions
- [ ] Field access
- [ ] Method invocation
I'm planning to submit PRs for each of these separately as and when I have figured them out.
This applies the same basic principle as #62062 to the reverse dataflow
analysis used to compute liveness information. It is functionally
equivalent, except that post-order is used instead of reverse post-order.
Some `mir::Body`s contain basic blocks which are not reachable from the
`START_BLOCK`. We need to add them to the work queue as well to preserve
the original semantics.
Currently, dataflow begins by visiting each block in order of ID
(`BasicBlock(0)`, `BasicBlock(1)`, etc.). This PR changes that initial
iteration to reverse post-order. This ensures that the effects of all
predecessors will be applied before a basic block is visited if the CFG
has no back-edges, and should result in less total iterations even when
back-edges exist. This should not change the results of dataflow
analysis.
The current ordering for basic blocks is pretty close to RPO
already--`BasicBlock(0)` is already the start block, so the gains from
this are pretty small, especially since we need to do an extra traversal
up front.
Note that some basic blocks are unreachable from the `START_BLOCK`
during dataflow. We add these blocks to the work queue as well to
preserve the original behavior.
This commit moves a test out of libstd which is causing deadlocks on
musl on CI. Looks like the recent update in musl versions brings in some
internal updates to musl which makes `setgid` and `setuid` invalid to
call after a `fork` in a multithreaded program. The issue seen here is
that the child thread was attempting to grab a lock held by a
nonexistent thread, meaning that the child process simply deadlocked
causing the whole test to deadlock.
This commit moves the test to its own file with no threads which should
work.
We've seen quite a few issues with spurious illegal instructions getting
executed on OSX on CI recently. For whatever reason `cc` itself is
executing an illegal instruction and we're not really getting any other
information about what's happening. Since we're already retrying the
linker when it segfaults, let's just continue to retry everything and
automatically reinvoke the linker when it fails with an illegal instruction.
For `libc` types that will be initialized in FFI calls, we can just use
`MaybeUninit` and then pass around raw pointers.
For `sun_path_offset()`, which really wants `offset_of`, all callers
have a real `sockaddr_un` available, so we can use that reference.
ci: Sync AppVeyor/Travis with Azure configuration
Manually make sure that we do the same thing across all the services,
uncovering one spot where we needed to pass one more configure flag on
Azure but otherwise we're good to go!
ci: Publish toolstate changes from Azure
This commit moves toolstate publishing from Travis to Azure. We've been
testing on azure for some time now and this works by deleting the Travis
config and updating the credentials used on Azure.
ci: Switch official `try` builds to happen on Azure
This commit switches the `try` builers to officially happen on Azure
Pipelines instead of Travis where they're currently run. This also cuts
back the number of builders to just the two we run on Travis, leaving
expansion as a possible future extension.
ci: Disable assertions in PR builds
The PR builder on Azure currently takes 2.5h which is a bit long, so
this commit disables debug assertions and llvm assertions in an attempt
to speed up that builder and have PR builds come back a bit more
quickly. Other builders continue to enable debug assertions and test the
compiler there.
Manually make sure that we do the same thing across all the services,
uncovering one spot where we needed to pass one more configure flag on
Azure but otherwise we're good to go!
This commit moves toolstate publishing from Travis to Azure. We've been
testing on azure for some time now and this works by deleting the Travis
config and updating the credentials used on Azure.
This commit switches the `try` builers to officially happen on Azure
Pipelines instead of Travis where they're currently run. This also cuts
back the number of builders to just the two we run on Travis, leaving
expansion as a possible future extension.
The PR builder on Azure currently takes 2.5h which is a bit long, so
this commit disables debug assertions and llvm assertions in an attempt
to speed up that builder and have PR builds come back a bit more
quickly. Other builders continue to enable debug assertions and test the
compiler there.
