On Fri, 2005-05-27 at 16:41 +0100, Keith MARSHALL wrote: > Bob Friesenhahn wrote, quoting me: > >> Yes, I see the logic of that. But, if configure has already > >> determined that the header file is not present, or at least not > >> usable, why would any user realistically want to do that? > > > > The autoconf philosophy is that the user (person who builds the > > software) should be in control. If some manual intervention from the > > defaults are required, it should not be necessary to edit files in > > order to handle that. > > > > The reason for overriding the existing/default configuration could be > > due to a poor choice by the configure script, or to experiment with an > > option without needing to re-run the configure script. > > Ok, that makes sense. Thanks. > > > Unfortunately, while the user is able to add definitions, I am not > > aware of a way to remove definitions other than to edit the configured > > header files. > > And, as Stepan has already noted, the ability to exclude definitions, > which configure has already added, could be potentially *more* useful > than adding those which configure thinks you shouldn't; IME, an autoconf > generated configure doesn't usually get this wrong! Well, ... you are neglecting ... * autoconf's history. It meanwhile "gets it right (tm)" more often than it did in the past. * compilers and systems gradually are converging and improving (towards c89/c99 and POSIX). This has gradually reduced the likelihood of hitting configure scripts producing bogus results. * Most configure scripts are comparatively simple. Complex configure scripts are comparatively rare and in most cases "don't drop out the blue", but gradually evolve over time, often over years. You'd probably be surprised, how often I have received reports of configure scripts, I maintain, "not getting it right" on "exotic systems" - In most cases the origin is not having considered something for such systems, in some cases it is autoconf (or automake, or libtool) "not getting it" right. In such cases, users sometimes (in rare occasions) resort to editing config-headers by brute force. Ralf _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
