Jan Nieuwenhuizen wrote: >>It seems to me that moving files (complete with dir structures) into >>separate $inst trees, just to make tha tar command simpler, is less >>than ideal. >> > > Why do you think that?
libpng12: mkdir -p $inst-libpng12/usr/bin mv $inst/usr/bin/cygpng-12.dll $inst-libpng12/usr/bin/ cd $inst-libpng12 tar cvjf libpng12.tar.bz2 * vs. cd $inst tar cvjf libpng12.tar.bz2 usr/bin/cyg*.dll You're assuming -- and it may be a safe assumption -- that all splitting will be done via one of two methods: 1) separate make install-* targets 2) whole subtrees I'm thinking more of ad-hoc splitting, where there are not separate make install-* targets, AND some files within a given $inst directory will end up in one package, and others from the same dir will end up in a different package. See my relatively recent post on how gettext was split up. What I'd need to do, under your scheme, basically use cd $inst tar -c -T $filelist -f - | tar -C $inst-other -x -f - to move the appropriate files, and then use tar AGAIN to create the actual archive. Kinda silly, BUT, having said all of that: I think your assumptions are probably safe, in general, and lead to a nice understandable structure in mknetrel's code. I can double tar if I need to. >>How does the existing mknetrel do it? (or does existing mknetrel have >>the ability to split up packages at all? >> > > I didn't think so. Did I overlook something? ??? I'll have to look... > Btw: you don't seem to have set a Mail-Followup-To: header, do you > mind (or actually like) getting all messages twice? Actually, my mail reader is smart. I only see one copy, even if I get two. --Chuck
