Michael Sperber <sperber(a)informatik.uni-tuebingen.de> wrote:
>>>>> "Stephen" == Stephen J Turnbull
>>>>> "ms" == Michael Sperber
ms> I find myself confused by the dichotomy between --enable-xxx
ms> and --with-xxx (especially as the configure script seems to
ms> know it's easy to get confused). Is there any easy way to
ms> predict when it's which one?
Stephen> Not as far as I can see. My taste differs from Malcolm's in several
Stephen> cases. AFAICT the implementations of --enable-xxx and --with-xxx are
Stephen> basically identical.
I don't know about the technicalities---but how hard is it to just
drop the ---with and --enable prefixes entirely? If not, I believe we
should just go with one of the two.
Normally, --enable should be used to activate internal features
(e.g. debugging), and --with should be used to compile stuff requiring
external components (e.g. png support). The frontier between the two is not
/that/ clear though.
In any case, if we're going to break backward compatibility, it shouldn't be
to invent yet another scheme of ours, but to conform more closely to the
autotools'tradition/convention. People using different autotools driven
packages like them to behave in a similar way (and will for instance issue a
quick "configure --help | grep --enable" to figure out what's available).
Didier Verna, didier(a)lrde.epita.fr, http://www.lrde.epita.fr/~didier
EPITA / LRDE, 14-16 rue Voltaire Tel.+33 (1) 44 08 01 85
94276 Le Kremlin-Bicêtre, France Fax.+33 (1) 53 14 59 22 didier(a)xemacs.org