Re: MinGw port of gawkextlib

[Eli Zaretskii]

> Date: Tue, 17 May 2022 12:41:22 -0400
> From: "Andrew J. Schorr" <aschorr@telemetry-investments.com>
> Cc: eliz@gnu.org, mcollado2011@gmail.com, bug-gawk@gnu.org
> 
> I think the simplest solution is to keep the existing configure option
> and add 2 more: --with-gawk-include-dir=DIR that tells us where to find
> gawkapi.h and --with-gawk-program=PATH. Does that cover all of our bases?

I don't think --with-gawk-include-dir=DIR is necessary, certainly not
for MinGW.  But if you prefer to have it, I don't mind, of course.

> But I think this is serious overkill. Nobody has ever complained about any
> of this before, most likely because nobody uses these extensions anyway.

Yes, and I guess very few tried to build the extensions with MinGW for
MS-Windows until now.

The question is: do you want to have gawkextlib extensions to have a
reasonable support of MinGW builds, including running the test suite?
If you do, then IMO you should provide a way for a user to tell the
configure script which Gawk executable to use for run-time tests,
because it is simply _impossible_ to run the test suite except by
using MSYS tools, and if MSYS is being used, the default Gawk is the
MSYS Gawk, which will never work for the test suite.  If such an
option were not available, I'd have to hack test/Makefile by hand to
force the test suite to invoke the correct Gawk.

IOW, the new --with-gawk-program=PATH option is required only for
running the test suite of the extensions, and only on MinGW, so you
could call it something like --test-with-gawk-program= or somesuch,
and document it as a MinGW only option.  Users on Posix platforms will
most probably never need that option.

> If the front-end installation script does its job properly, then this
> entire discussion will be moot.

Not sure I follow: how can an installation script solve the problem I
described above, which has to do with running the test suite of the
extensions?