Re: AC_CONFIG_SUBDIRS and subdirs

[Ralf Corsepius]

Am Mit, 2002-03-06 um 17.00 schrieb Akim Demaille:
> >>>>> "Ralf" == Ralf Corsepius <address@hidden> writes:
> 
> Ralf> Hi, I think, we've had this issue, before ;):
> 
> :)
> 
> Ralf> Alternatively, it should be clearly documented that that
> Ralf> AC_CONFIG_SUBDIRS does not support conditionally present subdirs
> Ralf> and users are supposed to check for presence of directories
> Ralf> themselves.  (Which, IMHO, would contradict the intention of
> Ralf> current autoconf's documentation on AC_CONFIG_SUBDIRS).
> 
> Well, maybe the doc ought to be adjusted, nevertheless, I think this
> is the correct feature.  Being able to detect an error seems better to
> me.  In addition, this is consistent with all the other AC_CONFIG.
> 
> So I don't plan to change this.  
Hmm. It is just inconsistent with handling of CONFIG_SUBDIRS at all
other places I am aware about. 

autoreconf treats them conditionally, _AC_OUTPUT_SUBDIRS does so, the
documentation does so, it's just AC_CONFIG_SUBDIRS which doesn't.

> Please, help us improve the doc, but
> if foo is optional, then the correct directives are
> 
> if test -f foo.in; then
>   AC_CONFIG_FILES(foo)
> fi
> 
> and
> 
> if test -d foo; then
>   AC_CONFIG_SUBDIRS(foo)
> fi
Well, I agree wrt. the restrictions which have been set upon
AC_CONFIG_FILES with autoconf-2.5x, but it is not _consistent_ with
regard to handling of config-subdirs at other places of autoconf.

But, config-subdirs have been documented to be optional. Therefore, I do
not understand why AC_CONFIG_SUBDIRS should not treat them optional.

Or to put it the other way round: Why should _AC_CONFIG_SUBDIRS,
autoreconf etc. treat them optional if AC_CONFIG_SUBDIRS doesn't?

IMHO, this logic is twisted/broken, unless a techical reasonale exits
for this behavior, which I currently don't see. (I am clearly in favor
of keeping config-subdirs optional).

Ralf