On Wed, 2004-05-05 at 11:58 -0400, Louis Guillaume wrote:

> Ronald van der Pol wrote:
> 
> > On Wed, Apr 28, 2004 at 14:41:31 -0400, Louis  Guillaume wrote:
> 
> >>Is there a way to speed up a  second run of "./build.sh -u release" 
> >>where the only reason the first run failed was because of flist problems?
> 
> > 
> > 
> >  I struggle with checkflist too
> > 
> > 
> 
> I've determined that the best way to do that "second run" is with the 
> "sets" target instead of "release". So the procedure for "extra files" 
> is - delete them manually, then do "./build.sh -u sets". It seems like 
> this results in a proper distribution. Any word to challenge that would 
> be helpful.
> 
> Regards,
> 
> Louis

I'm not sure if it is bad form to bypass build.sh, or include the extra
files but the script checkflist supports these arguments:

-x		check only x11 lists
-b 		check netbsd + x11 lists
-M metalog 	metalog file
-e		extra files are not considered an error
-m 		missing files are not considered an error

Which implied to me that forcing the sets to ignore errors can't be too
bad. So, I generate the failed sets with the following:

cd $SRCDIR/distrib/sets
$TOOLDIR/bin/nbmake-$ARCH sets CHECKFLIST_FLAGS='-e -m'

I'm crossbuilding so the above may be simplified if you're building
natively.

Alernatively you might be able to hand the CHECKFLIST_FLAGS through
build.sh using the -V switch. That might be a more correct way of
fudging it...

Tim