using librustc fails when trying to compile a second time

[jaredly]

The first time it works just fine, but it seems there's some global state somewhere that's messing with the parser. Here's the code: http://is.gd/K0dYsK

The main looks like:

fn main() {
    println!("Start");
    let src = "fn main() { println!(\"Awesome\"); }".to_string();

    let out = Path::new("/tmp/awesome".to_string());

    compile(src.clone(), out.clone(), None);

    println!("[[first worked]]");

    //this should work too, but it seems there's some state being held on to somewere, in the parser?
    compile(src.clone(), out.clone(), None);

    println!("[[second worked]]")
}

And the error is:

<anon>:1:1: 1:3 error: expected item, found `fn`
<anon>:1 fn main() { println!("Awesome"); }
         ^~
task '<main>' panicked at 'Box<Any>', /build/rust-git/src/rust/src/libsyntax/diagnostic.rs:86
playpen: application terminated with error code 101

[nrc]

cc @nick29581

[murarth]

I ran into this same issue not too long ago. There are two instances of thread-local data that are used in compilation. I can submit a PR for a fix, but the workaround is easy: As the data are thread-local, you can run each compile function in a separate task.

[nrc]

@murarth it would be great to have a fix if you have the time to put a PR together

[murarth]

@nick29581 No problem. I'm not sure about writing a test, though. How do I find sysroot from the test program?

[nrc]

@murarth I don't think you need it - you could have a makefile test which does something similar to @jaredly's example code and just check it doesn't panic

[jaredly]

Sweet thanks!

On Sat, Nov 29, 2014, 4:47 PM bors [email protected] wrote:

Closed #19371 #19371 via 8d8f41b
8d8f41b
.

—
Reply to this email directly or view it on GitHub
#19371 (comment).