Run rustdoc-gui tests in parallel

[dns2utf8]

I hid the passing tests and only show the failed ones in alphabetical order:

Also this PR cuts down the execution time from ~40 to ~9 seconds

[rust-highfive]

r? @Mark-Simulacrum

(rust-highfive has picked a reviewer for you, use r? to override)

[dns2utf8]

I extracted this from #86594 with @GuillaumeGomez

[GuillaumeGomez]

That might work well for linux, but not on windows at all. ;)

[dns2utf8]

I wished that JavsScript had #[cfg(unix)] ...

[GuillaumeGomez]

If that's the only thing you wished js had, I guess it's fine. 😉

For now, just keep it simple and simply print text.

[dns2utf8]

I found an easy way to detect linux at runtime and enable the colour, what do you think?

[GuillaumeGomez]

I think it's completely beyond scope. I'd rather keep it as simple as possible. Please simply print the text (on stderr if it's an error).

[dns2utf8]

Done 👍

[GuillaumeGomez]

Instead of putting everything into one string, it would maybe be better to keep an array of the tests and have some info for each, like output, errors, etc...

[GuillaumeGomez]

I also think we should disable running checks in parallel when --no-headless is used (take a look at src/test/rustdoc-gui/README.md to see how to use it).

[dns2utf8]

I implemented sequential execution if the --no-headless flag is present.

However, I was unable to make it run with ./x.py:

./x.py test --stage 1 --jobs 8 --no-headless --debug rustdoc-gui

Updating only changed submodules
Submodules updated in 0.03 seconds
    Finished dev [unoptimized + debuginfo] target(s) in 0.15s

Unrecognized option: 'no-headless'

Usage: x.py <subcommand> [options] [<paths>...]
...

[dns2utf8]

Regarding ordering the logs, I perallocate an Array now that the async tests fill with a result object. This can happen in order with the --no-headless flag or out of order without it.

The output then remains sorted with the exception of execution errors, which are at the end because I wanted people working with them to see them at the end of the output where they should be simpler to spot.
That just an idea I had and we can change it to just sorted output

[GuillaumeGomez]

I implemented sequential execution if the --no-headless flag is present.

However, I was unable to make it run with ./x.py:

./x.py test --stage 1 --jobs 8 --no-headless --debug rustdoc-gui

Updating only changed submodules
Submodules updated in 0.03 seconds
    Finished dev [unoptimized + debuginfo] target(s) in 0.15s

Unrecognized option: 'no-headless'

Usage: x.py <subcommand> [options] [<paths>...]
...

Look here how to pass arguments to the tester through x.py.

[GuillaumeGomez]

Shouldn't it be:

Suggested change

	await Promise.all(tests);
	if (!no_headless) {
	await Promise.all(tests);
	}

?

[dns2utf8]

Logic wise kind of yes. But you can await a future multiple times, the work will get done only once

[GuillaumeGomez]

Shouldn't it be:

Suggested change

	console.log("\n");
	console.log("");

? Or do you specifically want an empty line?

[GuillaumeGomez]

If so, maybe better to use two console.log("") instead I think. No strong opinion though.

[dns2utf8]

I wanted a bit more separation between the sections for readability so the progress bar is visually spaced from the outputs.
The output should be the same on all platforms because we don't write anything on the empty line but fully correct it should be two console.log(""), will change that

[dns2utf8]

Replaced it with a better finish method that prints the final stats

[GuillaumeGomez]

A few questions to be answered but otherwise it looks really promising!

[dns2utf8]

I am using os.EOL now for the line endings. Also the status_bar now has a finish function that aligns the last progress bar.

Here is an example run with two failed tests and --debug:

And here is one without:

One small question for me is how many characters do we want per line? Currently I picked 10, but this is arbitrary and maybe a bit low?

[GuillaumeGomez]

One small question for me is how many characters do we want per line? Currently I picked 10, but this is arbitrary and maybe a bit low?

Other test-suites use 100 but I don't care. We can always update it later. Looks all good to me now, thanks a lot!

@bors: r+

[bors]

📌 Commit 7f2b52b has been approved by GuillaumeGomez

[bors]

⌛ Testing commit 7f2b52b with merge 58d685e...

[Mark-Simulacrum]

Is this respecting -j passed to x.py somehow? We should avoid running multiple processes if -j1 is passed, for example.

[bors]

☀️ Test successful - checks-actions
Approved by: GuillaumeGomez
Pushing 58d685e to master...

[GuillaumeGomez]

@Mark-Simulacrum Good point. I'll open an issue for it.