* Tests are run in order of their filenames (files first, then recursing into folders).
So if you have any slow tests, prepend them with a small integral number to make them get run first, taking advantage of parallelism as much as possible (instead of waiting for the slow tests at the end).
If your test tests for failure, you need to add a `//~` annotation where the error is happening
to make sure that the test will always keep failing with a specific message at the annotated line.
`//~ ERROR: XXX` make sure the stderr output contains `XXX` for an error in the line where this comment is written
* Also supports `HELP`, `WARN` or `NOTE` for different kind of message
* if one of those levels is specified explicitly, *all* diagnostics of this level or higher need an annotation. If you want to avoid this, just leave out the all caps level note entirely.
* If the all caps note is left out, a message of any level is matched. Leaving it out is not allowed for `ERROR` levels.
* This checks the output *before* normalization, so you can check things that get normalized away, but need to
be careful not to accidentally have a pattern that differs between platforms.
* you can specify this multiple times, accumulating all the env vars
*`//@normalize-stderr-test: "REGEX" -> "REPLACEMENT"` replaces all matches of `REGEX` in the stderr with `REPLACEMENT`. The replacement may specify `$1` and similar backreferences to paste captures.
* you can specify multiple such commands, there is no need to create a single regex that handles multiple replacements that you want to perform.
*`//@require-annotations-for-level: LEVEL` can be used to change the level of diagnostics that require a corresponding annotation.
* this is only useful if there are any annotations like `HELP`, `WARN` or `NOTE`, as these would automatically require annotations for all other diagnostics of the same or higher level.
