Check fails, stdout and stderr

[projetmbc]

When check fails, you don't use the stdout stream. It should be better to indicate the failure "Check failed with <>" inside the stderr stream. Doing this allow extra tools to catch only checking that gives an error without having to parse all the stdout stream.

[josephwright]

This is not an error from the l3build point of view: it's information for the user.

[projetmbc]

This makes sense. But I see more that as a workflow error than a computation one. This makes me write weird lines of code : ugly tests .

[josephwright]

I suspect you are doing something ... odd from our point of view: the entire idea with l3build is you run it yourself then take appropriate action if there is a failure. Could you outline your use case?

[projetmbc]

Using parallelism, my script speeds up the search for regression when I modify my code. If I have an unexpected failure, I then work on each failing test on a case-by-case basis.

PS: in the end, once everything is fixed, I use l3build ctan, and therefore a classic l3build check.

[josephwright]

Using parallelism, my script speeds up the search for regression when I modify my code. If I have an unexpected failure, I then work on each failing test on a case-by-case basis.

We do that too :)

[josephwright]

@zauguin You are the expert in this area - what's your take?