On 01/24/11 10:31, Eric Blake wrote: > Hmm, it may still make sense to do both things - make it possible > to completely elide HAVE_RAW_DECL_* (this patch) for configure > built for a tarball, as well as to provide a configure option > (rather than the current ad-hoc approach of > make CFLAGS=-DGNULIB_POSIXCHECK=1) to make it more apparent when > a developer is explicitly building for posix check purposes > and to thus speed up builds when the option is not in effect.
That sort of thing sounds useful, yes. I have another idea or two which you may want to throw into the pot. When I used gl_ASSERT_NO_GNULIB_POSIXCHECK in Emacs (see <http://git.savannah.gnu.org/cgit/emacs.git/commit/?id=8d38eb490bdd94432de803575ac3663cef360285>) aclocal helpfully informed me that m4/warn-on-use.m4 was unnecessary, so I added a line "rm m4/warn-on-use.m4" to the Make rule for "make sync-from-gnulib". This is a bit ad-hoc and it'd be nice if gnulib-tool would automate it. More generally, it'd be nice if there were a package "no-gnulib-posixcheck" that would do all this for me. Or, perhaps even better, a package "gnulib-posixcheck" that would enable the HAVE_RAW_DECL_* stuff, with the default being to disable it. The latter might be easier because then its Files: section could list m4/warn-on-use.m4, and we could remove m4/warn-on-use.m4 from the warn-on-use module, and remove the need for gl_ASSERT_NO_GNULIB_POSIXCHECK.
