Re: gettext patches for cygwin #2: autoconf warnings, cygwin/getopt

[Charles Wilson]

Bruno Haible wrote:

> Actually, the gettext testsuite is buggy. I've now made the testsuite
> work with "env POSIXLY_CORRECT=1 make check". So it is not needed any
> more to use GNU getopt if the system already has a fully POSIX compliant
> getopt() and getopt_long().
>
> If a system does not support the GNU extension which allows to put
> option arguments after non-option arguments, it's a choice of the system's
> architects. (It has an effect on users, who must tediously port their
> shell scripts, reordering arguments in command invocations. But on the
> standards side, there's nothing to say against it.) GNU gettext follows
> this choice of the platform architects, by using the platform-provided
> getopt().

Well, it's probably a good thing that you did that, but not for cygwin. 
 Turns out my info is out of date; I'm sorry for the confusion.  Cygwin 
replaced its internal version of getopt with the vanilla BSD version 
(more or less) about 18 months ago.  So, POSIXLY_CORRECT isn't hardcoded 
to 1 at all anymore (actually, the very very old cygwin getopt didn't 
have PC at all; it just behaved as the generic getopt would have done, 
IF PC were 1.  A slighly newer but still ancient version in the cygwin 
CVS history DEFAULTED to PC=1 behavior, but could be overriden with [get 
this!] env var POSIXLY_INCORRECT_GETOPT).

Anyway, I've been carrying my patches around in the cygwin releases of 
gettext since around 2001...and I've never checked to see if they were 
no longer necessary, until now.

It's likely that the getopt-related patches could have been dropped from 
my cygwin-gettext build last Feb with no ill effect.

Again, sorry for the noise.  But what of the AH_BOTTOM/AH_TOP 
replacements, to quiet the autoconf deprecation warnings?

--
Chuck