Re: GNATS needs some work (autoconf/59)

[Pavel Roskin]

Hello, Tim!

> I'd like to volunteer to handle anything DOS/DJGPP-related if
> possible.

Thanks!

> As far as bug 59 (the pathsep thing) goes, I'm already looking
> into a good way to do this. The way I see it, there are two
> different issues that need to be handled separately.

That's very insightful.

> On the one hand, there is the pathsep on the build system, which
> should be used by autoconf, automake-generated makefiles and
> other build scripts (such as autoconf's testsuite). Then there
> is the pathsep on the target, as would be AC_DEFINE'd for use in
> a program.

It's not called TARGET. It's called HOST. TARGET is only meaningful for
compilers and other tools working together to create executables _for_ a
particular platform.

HOST is where your program will run. It's meaningful for all programs.

I'd like to advise you that it would be fair not to support non-UNIX path
separators on the build system. Autoconf doesn't pretend to support such
systems.

On the other hand, support for non-UNIX path separators on the host would
be a feature that very few people need. It falls outside the area of
"generic" macros that Autoconf contains.

A notable example would be Kaffe or any other Java implementation that
tries to mimic the existing Java tools for that platform. But that's
clearly too specific to be part of Autoconf.

Ideally, you shouldn't copy or require config.guess. Try to test the
behavior, not to guess it. For example (not tested!):

mkdir constest.dir
cat > constest.dir/constest.bat <<EOF
echo foo
chmod +x constest.dir/constest.bat
for path_sep in ":" ";" ; do
  PATH=constest.dir${path_sep}$PATH
  if test `constest.bat 2>/dev/null` = foo; then
    break
  fi
done

Again, if it fails or will be too hard I'll consider it fair to drop this
idea.

Regards,
Pavel Roskin