[m-users.] [mercury-users] .mh files

[Volker Wysk]

Am Mittwoch, dem 24.07.2024 um 12:09 +0200 schrieb Zoltan Somogyi:
> 
> On Wed, 24 Jul 2024 11:53:00 +0200, Volker Wysk <post at volker-wysk.de> wrote:
> > > > I never needed to open a .err file. I'm using "--make" and the compiler
> > > > outputs all error messages...
> 
> That is not actually true. The compiler outputs its diagnostics to the .err file,
> and then copies an *initial subset* of those diagnostics to its own output stream.
> However, unless you specify --no-output-compile-error-lines, some diagnostics
> will NOT appear in the output of mmc --make.

Okay. So the compiler outputs all *requested* messages - requested either
implicitly or by explicit command line option.

> > > Yes, I would prefer to put .err files in a subdir whenever --use-subdirs
> > > or --use-grade-subdirs is in effect (that includes mmc --make).
> > > 
> > > Users using mmake *do* need to open .err files, but mmake doesn't use
> > > subdirs by default, so maybe that is fine.
> > 
> > Yes, it has to be good for something, or it wouldn't be this way. It's just
> > that users of "--make" don't need those .err files.
> 
> Some users of mmc --make *do* need .err files. One reason I mentioned above.
> Another is that some people use automatic error-message-processing programs 
> such as extras/error. 

In this case, I would prefer a command line option that triggers the
creation (or retention) of the .err files. Or, they also could be placed in
the "Mercury" directory, like Peter has suggested. That seems to be the best
option to me.

> A third is the Mercury test suite, which needs to compare actual
> compiler-generated .err files with their expected contents.

Dito.

> Leaping from "I don't need this" to "nobody can possibly need this" is not
> a good idea.

Okay. I weaken my statement to "Many users don't need this".

I just want to give some feedback.

Cheers,
Volker