[m-dev.] errors from mmc script when compiling srcdist

[Peter Wang]

On Wed, 18 Oct 2023 15:22:49 +1100 Peter Wang <novalazy at gmail.com> wrote:
> On Wed, 18 Oct 2023 14:46:29 +1100 Julien Fischer <jfischer at opturion.com> wrote:
> > 
> > Hi,
> > 
> > Building the source distribution for rotd-2023-10-16 onwards results in
> > a lot of messages from the mmc script about mercury_compile not being
> > found (which it wouldn't be as it doesn't exist when compiling the
> > source distribution).
> > 
> > I suspect some of the recent changes to mmake may have inadvertently
> > broken something.
> > 
> 
> Reverting commit 7d001786 fixes the issue. I'll have a look.

GRADESTRING and GRADESTRING_NO_TARGET both call $(MCOGS)
as a shell command, where

MCOGS           = $(MC) --output-grade-string

but mercury_compile hasn't been built yet. How did this ever work?
Presumably when VPATH used to indirectly refer to GRADESTRING,
the fact that $@ and $* were undefined caused make not to run the
shell command.

I think we can check if EXTRA_LIB_DIRS is the empty string and,
if so, assign EXTRA_INIT_DIRS_NO_TARGET to the empty string
to avoid the patsubst that refers to GRADESTRING_NO_TARGET.
I'll send a patch shortly.

Peter