Re: 'wait -n' with and without id arguments

[Greg Wooledge]

On Fri, Aug 09, 2024 at 13:59:58 -0400, Zachary Santer wrote:
> I don't necessarily understand why someone would call 'wait -n' from
> the interactive shell, so I don't really know what the desired
> behavior would be when they do so. Would be nice if other people want
> to chime in on that point.

The only use case I'm aware of for "wait -n" was implementing a thing
that runs N processes at a time, launching one new process every time
an existing one completes.

I could easily see someone writing a function for this and putting it in
their .bashrc and using it from an interactive shell, though I haven't
done this myself.

The problem is that our entire understanding of what "wait -n" DOES has
been annihilated.  We thought it would "trigger" exactly once for every
completed background process, regardless of whether they completed
before or after calling "wait -n", which would allow the writing of an
N-jobs-at-a-time thing.  It turns out this is incorrect.

Now that we know "wait -n" is a gigantic race condition with no
predictability, it's hard for me to come up with any legitimate use
for it.

The best N-jobs-at-a-time implementation now is probably GNU xargs -0 -P,
but it's a VERY distant second place behind the mythical wait -n loop.