RE: AC_PROG_INSTALL fix

[Tim Van Holder]

> Thanks for the patch, but I laready have something similar at hand,
> and you are still burned :(
You mean that I am on your blacklist, or that my papers _still_ haven't
arrived?

> Anyway, your patch is weird: you break the IFS
How so? See below, I just copied it from AC_SHELL_PATH_WALK. In fact
keeping IFS set to ':' (or ';') for the duration of the look breaks
the expansion of variables whose values contain spaces.

> you add a \ where none seem to be needed,

> you introduce an $ac_dummy whcih seems as useless. ($PATH
> is split just like nay other variable, why would you need another one?)
AC_PROG_INSTALL was not finding install, and investigation suggested
IFS had something to do with it. So I simply copied the relevant bits from
AC_SHELL_PATH_WALK:

m4_define([AC_SHELL_PATH_WALK],
[ac_save_IFS=$IFS; IFS=$ac_path_separator
ac_dummy="m4_default([$1], [$PATH])"
for ac_dir in $ac_dummy; do
  IFS=$ac_save_IFS
  test -z "$ac_dir" && ac_dir=.
  $2
done
])

->

AC_MSG_CHECKING([for a BSD compatible install])
if test -z "$INSTALL"; then
AC_CACHE_VAL(ac_cv_path_install,
[  ac_save_IFS=$IFS; IFS=$ac_path_separator
  ac_dummy="$PATH"
  for ac_dir in $ac_dummy; do
    IFS=$ac_save_IFS
...

If the ac_dummy is superfluous in AC_PROG_INSTALL, then why is it used
in AC_SHELL_PATH_WALK?

> If this is needed, then we really need some more information, and comments
> in the code.
Don't see why - aside from the unnecessary backslashes, the changes were
taken from an existing, working macro to fix a macro that was not doing its
job.