Potential compiler memory leak in Nightly

[sazzer]

(Apologies if I've done this wrong. I've used the Bug template and not the ICE one because I'm not getting the "Internal Compiler Error" error message. I've also had a bit of a search and not seen anyone else reporting this at all, but I can't see anything unusual that I'm doing.)

I have a project that recently stopped building, and it's taken me a while to work out why. I still don't know everything - sorry! - but what I do know is this:

• I'm using Nightly. It does not affect Stable. Additionally, it does not affect Nightly from before 2020-08-13. The very first version that breaks is:

-> % rustc --version --verbose
rustc 1.47.0-nightly (81dc88f88 2020-08-13)
binary: rustc
commit-hash: 81dc88f88f92ba8ad7465f9cba10c12d3a7b70f1
commit-date: 2020-08-13
host: x86_64-apple-darwin
release: 1.47.0-nightly
LLVM version: 10.0

• I'm getting this on macOS but it also affects Linux. I know because I first caught it on my Github Actions CI build.
• When it happens, I can got to Activity Monitor and see the memory usage of rustc going up. I caught this one because the OS complained about running out of memory on one test when it reached 15GB of memory used by the compiler.
• I'm using a Cargo Workspace to keep different components in my system isolated from each other. However, I can't see any reason why that would be the problem. The error I'm seeing happens deterministically in exactly one of the crates, and all of the others are fine.
• I have one single line of code that I can comment out that will stop the problem happening. However, it's not anything special.

The error happens in the method:

#[get("/authentication/{provider}/complete")]
#[tracing::instrument(
    fields(http_method = "GET", http_path = "/authentication/:provider/complete"),
    skip(authentication_service, authorization_service)
)]
pub async fn complete(
    path: Path<(ProviderId,)>,
    query: Query<HashMap<String, String>>,
    authentication_service: Data<AuthenticationService>,
    authorization_service: Data<AuthorizationService>,
) -> HttpResponse {
    let user = authentication_service
        .complete_authentication(&path.0, &query.0)
        .await;
    todo!()
}

If I remove the .await call in there then the build completes. If I leave the .await in then it breaks.

However, I've got other Actix endpoints that use async/await and they work perfectly well.

Cheers

[jyn514]

@sazzer it would be great if you could reproduce this in a stand-alone example, or at least the commit of the project that causes the hang.

[sazzer]

@jyn514 The project link would have helped, wouldn't it? Sorry! :(

• This is the Github Actions report for the failure: https://github.com/sazzer/mire/runs/983949452?check_suite_focus=true
• This is the actual branch that has the problems: https://github.com/sazzer/mire/tree/rust

(There are some other branches that I've been playing with other things on. It's the "rust" branch that is interesting here)

When doing either cargo test or cargo build it will go fine until you get to "mire-authentication" and that's the one that hangs. If you edit crates/authentication/src/endpoints/complete.rs to remove the .await on line 32 onwards then the build will work fine.

(Please don't judge me on the quality of the code - I'm still learning :) )

Cheers

[jyn514]

Thanks! Don't worry about the code quality, part of rustc's job is to give good errors for broken code, not just code that's already working :)

(the commit that failed was https://github.com/sazzer/mire/commit/0bfad66a070673c58c1e98d334c5500712c77e63)

[sazzer]

Just to clarify btw - it's not the commit that fails. That commit works fine on an earlier nightly toolchain, and on the stable toolchain. I've also not seen which of the earlier commits also fail on the newer nightly toolchain - though I can do if it would be useful? I suspect that it starts from https://github.com/sazzer/mire/commit/589c1b7.

[sazzer]

Ok - the first commit that shows the problem appears to be https://github.com/sazzer/mire/commit/c927b77.

[LeSeulArtichaut]

It would be great to do some bisection and try to find an MCVE.
@rustbot ping cleanup

[rustbot]

Hey Cleanup Crew ICE-breakers! This bug has been identified as a good
"Cleanup ICE-breaking candidate". In case it's useful, here are some
instructions for tackling these sorts of bugs. Maybe take a look?
Thanks! <3

cc @AminArria @camelid @chrissimpkins @contrun @DutchGhost @elshize @ethanboxx @h-michael @HallerPatrick @hdhoang @hellow554 @imtsuki @kanru @KarlK90 @LeSeulArtichaut @MAdrianMattocks @matheus-consoli @mental32 @nmccarty @Noah-Kennedy @pard68 @PeytonT @pierreN @Redblueflame @RobbieClarken @RobertoSnap @robjtede @SarthakSingh31 @senden9 @shekohex @sinato @spastorino @turboladen @woshilapin @yerke