2015-01-07 19:14 GMT+03:00 Florian Stinglmayr <flor...@n0la.org>: > On Wed, Jan 07, 2015 at 04:41:16PM +0100, David Coppa wrote: >> On Wed, Jan 7, 2015 at 4:36 PM, Stuart Henderson <st...@openbsd.org> wrote: >> > >> > Hmm. Actually what is the benefit of using >> > CMAKE_SKIP_INSTALL_ALL_DEPENDENCY >> > anyway, even in the make(1) case? >> > >> >> It was requested by Vadim some years ago: >> >> "This will tell CMake to skip rescanning the whole directory in the >> fake stage. Our ports infrastructure already takes care that fake >> happens only after build, and this re-scanning is very annoying >> when you are debugging something in a somewhat fat package. >> And this change should speed up bulk builds too..." >> > > This could also be achieved by calling the target "install/fast" instead > of "install" in the fake stage. "install/fast" also causes cmake to skip > all dependency checks.
This sounds like something requiring bulk build to check. Is there any real gain (e.g., difference with current situation)? Also, I don't see any problem with having -DCMAKE_SKIP_INSTALL_ALL_DEPENDENCY until all ports on all archs stop using make backend. David, what do you want to achieve (except shrinking cmake.port.mk, which is nice itself, of course)? > On the other hand, a new "if" just for cmake is > probably not the cleanest solution either. -- WBR, Vadim Zhukov