On Thu, 2004-07-22 at 00:55, Toshio wrote: > On Wed, 2004-07-21 at 17:45, Ralf Corsepius wrote: > > On Wed, 2004-07-21 at 19:08, Toshio Kuratomi wrote: > > > I too think that the "proper" way to deal with problems with auto*tools is > > > to patch. However, I think it's impractical. Not from size of the src.rpm, > > > but from size of the patch. It is a nightmare to properly QA the tangled > > > patch of regenerated Makefiles, Makefile.ins, configure, et al. > > > > Working around this topic is simple - Split the diff into two: One > > containing the patches to the sources (configure.acs, Makefile.ams) and > > one patch containing the generated files. > > > Huh? I regularly do that. The sources patch remains nice and small. > But the generated files patch is still huge. I was referring to your remark on QA'ing auto*generated files, not on the size. > If I was QA'ing this package, I'd be able to check out the base patches > and spec relatively easily but the postautogen patch would be quite a > chore. I'd browse them and check for something catching my eye. Otherwise I'd treat them more or less as "binary" and trust in the person who has generated them. Of cause there is a potential risk of somebody trying to hide something malicious in such a patch, but ... he'd only do that once ;-) I'd be more concerned about the person not having generated the files correctly or the person having used incompatible versions of the autotools (cf. ethereal discussion a couple of days ago). These problems are easier to catch when actually generating the files inside of the spec. Ralf
