[m-rev.] for post-commit review: fix decision about parallel folds

[Julien Fischer]

On Tue, 27 Aug 2024 at 12:35, Zoltan Somogyi <zoltan.somogyi at runbox.com> wrote:
>
>
>
> On Mon, 26 Aug 2024 17:25:55 +1000, Peter Wang <novalazy at gmail.com> wrote:
> > I think the idea was that failing to set up the IPC for a parallel fold
> > is unexpected, so we should stop ASAP.
>
> I also think that is a good idea, but only if we report the actual problem
> as the reason we stopped. The code the diff replaced reported the problem
> as being with the program being compiled, which is very misleading.
> Unfortunately, the make package's infrastructure has no way to distinguish
> between "did not succeed due to a platform problem" and "did not succeed
> due to an error in the program being compiled"; it assumes all failures
> are of the latter kind.
>
> We *could* improve the infrastructure to allow us to make this distinction.
> The question is: is this kind of failure frequent enough to make that programming
> effort worthwhile? I don't use mmc --make myself, but my guess is that the
> answer to that is a resounding "no".

I have never seen parallel mmc --make fail due to an IPC problem.

Julien.