[m-dev.] testing.mercurylang down?

[Zoltan Somogyi]

On Sun, 13 Oct 2024 16:03:33 +1100, Julien Fischer <jfischer at opturion.com> wrote:
> Oh, I see what's happened -- previously we would have just overwritten
> each Deep.data
> and Deep.procrep file but since they no longer have the same name that
> no longer happens.
> We have mechanisms to disable the generation of deep profiling files,
> but in .profdeep
> bootchecks we should be writing them out IMO (i.e. to test that
> writing them out works).

Actually, we don't write them *all* out; we write out a sort-of-random sample.
I seem to recall that we write out the profile only about 4% of the time.
Unless a bootcheck is very unlucky, this should still find all the bugs
that exist, but significantly reduce both the time and the space cost.

> I suppose we should add a runtime option that lets us use the old
> Deep.data and Deep.procrep
> names (or some other standardized names).

I can do that later today, unless someone has already started on it.

Zoltan.