On Tue, 2002-09-24 at 01:09, Mark McLoughlin wrote: > Hey, > > On 23 Sep 2002, Bastien Nocera wrote: > > > On Mon, 2002-09-23 at 23:08, Thomas Vander Stichele wrote: > > > I've been meaning to propose that too. > > > > > > While you're at it, AFAIK you can do the same for autoconf, autoheader and > > > automake. If they return a value, the problem is almost always fatal and > > > hard to spot based on further build or compilation logging. Unless > > > someone knows of specific cases where any of these exits with non-zero but > > > the error is actually recoverable and not easily avoidable ? > > > > > > Thomas > > > > > > > Heya, > > > > > > > > I'm getting funky errors from people who don't install the ALSA libs and > > > > want to compile ACME CVS. > > > > > > > > Here's a patch to fail autogen.sh when aclocal-1.4 fails. > > > > > > > > OK to commit ? > > > > Dunnit. > > > > Patch attached. Works here. OK to commit ? > > A lot of the time auto* non-zero returns *are* non-fatal > errors. Exactly, they're not fatal when they're run. But they cause fatal errors later in the building process. My whole point. > Committing this patch would be equivalent to foisting -Werror > on the whole stack. At the very least branch a gnome-2-0 branch of Huh, except that this fix works with all C compilers. > gnome-common before committing so that the resulting build breakage is > at least only on HEAD. I don't really see the point here. It's a fairly trivial patch indeed, and it does what it should always have done. I have tested quite a few modules with it and it works. Anybody who does complete CVS compilations fancy testing ? Cheers -- /Bastien Nocera http://hadess.net
Attachment:
signature.asc
Description: This is a digitally signed message part