Move remaning compile-fail tests that are rejected by the parser to pars...

[fhahn]

After making rustc fail on errors at a stop point, like -Z parse-only, in #22117, the files in this PR also fail during the parse stage and should be moved as well. Sorry for spliting this move up in two PRs.

[rust-highfive]

r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

[alexcrichton]

Hm I don't actually think we're running any parse-fail tests at all! Looking at the most recent output for a successful test run there are no mentions of any parse-fail tests. I think that this change was only meant for re-running compiletest, the actual crawling of a directory comes from the --src-base argument and the logic may need to be tweaked to have "two source bases" for this.

Can you update this PR as well to be sure to run all the just-moved parse-fail tests and make sure that make check runs all tests?

[Manishearth]

You might want to add Manishearth@19d8d3e to this PR too.

[fhahn]

@alexcrichton damn, sorry! Should I modify compiletest.rs to accept more than a single path as --src-base or should I add a separate make target for parse-fail?

[alexcrichton]

Whatever is easiest is fine by me, I'd probably go with multiple --src-base arguments as I think getopts already handles it. Also no worries, don't sweat it!

[fhahn]

I've opted to add another pfail target, because I was not sure how to add multiple --src-base arguments in the Makefile in a generic way. Also note that at the moment, I am using the CompileFail mode of compiletest, so running the pfail tests prints [compile-fail] parse-fail/..... This may seem confusing, should I change that?

However I am not sure which part of the Makefile is responsible for copying the test files to the test directory. Any idea what I could have missed or pointers where I could start looking?

[alexcrichton]

When you say copying test files, what are you referring to? I think your changes here should be sufficient for running the tests.

[fhahn]

I got it wrong, I just had to create the ´test/parse-fail´ directory in the configure script. It should run the parse-fail tests now. But compiletest still prints [compile-fail] parse-fail/...., because I am reusing this code. Could this be somehow confusing?

[fhahn]

@Manishearth the test case you reference fails during macro expansion (or later, I am not sure), because the include! has to be expanded first. So I think adding it to parse-fail would not be such a good idea, because I would like to add the -Z parse-only option to all parse-fail tests and the LALR parser does not handle macro expansion (although we could include some extra code to also parse files included with include!)

[brson]

@fhahn I would prefer if you could add a ParseFail mode to compiletest if it's easy to do.

[fhahn]

@brson I've updated the code.

[alexcrichton]

ccheck typo

[alexcrichton]

(ping typo)

[alexcrichton]

Thanks @fhahn, looks good to me! Just a small typo and otherwise r=me.

Also, just to confirm, but you see the output of the tests in make check?

[fhahn]

@alexcrichton I have fixed the type and now parse-fail and compile-fail tests show up when running make check

[alexcrichton]

(I think the typo is still there)

[fhahn]

Argh yes, I somehow forgot to push the updated commit. Sorry for that!

[alexcrichton]

@bors: r+ 6824f13

(I'll fix the typo in the upcoming rollup)