In article <20130315130901.gd...@pony.stderr.spb.ru>, Valery Ushakov <u...@stderr.spb.ru> wrote: >On Fri, Mar 15, 2013 at 16:56:53 +0400, Valery Ushakov wrote: > >> On Thu, Mar 14, 2013 at 21:02:06 -0700, Tom Spindler (moof) wrote: >> >> > > > >> >Log Message: >> > > > >> >Revert 1.25 and 1.26: Breaks build on !NetBSD >> > > > >> >> > > > >> Can you explain how? The shell does not work? >> > > > > >> > > > >E.g. util.h not found in binstall and a lot of other header issues. >> > > > >> > > > On what os? >> > > >> > > I've seen it on OSX ( 10.5 PowerPC if it matters ). >> > >> > And I've seen it on Debian. >> >> That begins to sound like "I've got it on MacOS", err, sorry, "I've >> got it from Agnes" song. :) >> >> Was there an error message, perhaps? > >Ah, as Martin pointed out privately, the problem is not some bug in >the change ("Can you explain how? The shell does not work?"), but the >fact that a missing include is now fatal, which is a wrong thing to do >for mkdep.
I still don't understand how a missing include does not cause an error in the compilation phase, but it can be ignore in the dependency generation. christos