[m-rev.] request for test cases: libdir_map in make.library_install.m

[Julien Fischer]

On Fri, 23 Aug 2024 at 03:22, Zoltan Somogyi <zoltan.somogyi at runbox.com> wrote:

> On Thu, 22 Aug 2024 23:45:43 +1000, Julien Fischer <jfischer at opturion.com> wrote:
> > > The updated diff is attached. With it, an install of the complex_numbers
> > > library works.
> >
> > With the mli3a diff --use-grade-subdirs does not work.  Building samples/diff
> > using:
> >
> >     mmc --use-grade-subdirs --make diff
> >
> > fails with:
> >
> > $ mmc --use-grade-subdirs --make diff
> > Making Mercury/int3s/diff.int3
> > ** Error reading file `builtin.m' to generate dependencies.
> > ** Module `builtin' is imported by module `diff'.
> > mercury_compile: cannot find source for module `builtin' in directories `.'
>
> Sorry, I forgot to mention that this diff was not intended to fix
> that issue, which is what Peter reported. It was intended only to
> clarify, and thus let me to understand, the code that I would have
> to modify to fix that bug :-(

Yes, I somewhat belatedly realised that.

> > (G12 doesn't build since its build system requires the use of
> > --use-grade-subdirs.)
>
> OK. Since I don't know of any libraries more complex than
> complex_numbers, and it seems that you guys don't have any either
> that work without --use-grade-subdirs, I would say that this diff
> is now as tested as it is going to be.

I now have G12 building without --use-grade-subdirs and can report
that it builds, installs and passes its regression tests using the mli3a
diff.  (I reviewed the contents of the install directory and cannot see
anything amiss there.)

Julien.