<quote author="Keith Marshall>
>> # MSYS_PREFIX_HACK
>> # If the user didn't specify any prefix,
>> # AND the build system supports `pwd -W`, (i.e. is probably MSYS),
>> # set the default autoconf prefix, as a Win32 format path.
>> # Has no effect, UNLESS the build system supports `pwd -W`,
>> # AND '/usr/local' exists in the MSYS pseudo file system.
>> [if test "$prefix" = "NONE"; then
>> prefix=`exec 2>/dev/null; cd /usr/local && pwd -W` ||
>> in aclocal.m4, (or some site specific autoconf macro file), and to
>> it via configure.ac, when configure is generated.
>> What do you think? I may propose something along these lines, for the
>> groff project.
<quote author="Earnie Boyd">
> Well, it should be MINGW32_PREFIX_HACK, IMO; but yes I think it should
> the way to go.
Whatever -- the name doesn't much matter to the functionality. I thought
MSYS_PREFIX_HACK, because it relies on a feature of the MSYS shell, and is
rather compiler neutral. It really wouldn't help users of MinGW, who
use the MSYS environment for running their build scripts, unless of
their chosen build environment uses POSIX file system emulation similar to
that provided by MSYS, *and* their shell also supports the 'pwd -W'
Maybe there is a case for keeping this test as MSYS_PREFIX_HACK, and also
providing a complementary MINGW32_PREFIX_HACK macro, which would rely on
an AC_COMPILE_IFELSE, or an AC_RUN_IFELSE type test.
Just a further thought -- if this is the way to go, wouldn't it be nice if
it was available as a standard autoconf macro? Perhaps we should suggest
it to the autoconf maintainers, (so it would become AC_MSYS_PREFIX_HACK,
or AC_MINGW32_PREFIX_HACK, if you prefer).