On Wed, Sep 15, 2010 at 9:34 AM, Paul Eggert <egg...@cs.ucla.edu> wrote: > On 09/15/2010 08:41 AM, Sam Steingold wrote: >> I have to reiterate my complaint that the gnulib dependencies mean that one >> has >> to include (almost) the whole of gnulib if one includes (almost) any part >> thereof. > > I agree that gnulib has too many dependencies and [...]
Once upon a time, long ago and far away, I proposed a new library, "libcompat" that more-or-less would make all target platforms look pretty much identical. Lo and behold, over the past decade, gnulib has gradually grown up to provide all that infrastructure. The only problem with it is that gnulib is glued in to each project adding truly remarkable amounts of configure time overhead to each and every project that depends upon its glue, and that would be true even with minimal gnulib interdependency. I do think it would be Really, Really Nice if there were a gnulib.pc file one could test. If it were not there, I could just fail the configure of my package and forget the several minutes of configuration overhead. If there were some recent gnulib feature I could not live without, then I might add that one bell or whistle to my project and the configure process might be extended by a fraction of a second. So, let's start that discussion again: let's make an installable gnulib that basically does all the configure tests for the current target and cover 99% of all configure runs. If someone has a strange configuration with extra libraries et al., then it ought to be straight forward enough to have an alternate gnulib.pc file installed in an alternate PATH setup. I do not see an obvious reason why the configure process should be so ridiculous. Or is that just me? ;)
