Le mardi 24 novembre 2009 02:00, Yasushi SHOJI a écrit : > Hi all, > > it seems a classic case. > > At Mon, 23 Nov 2009 16:41:44 -0800, > Greg KH wrote: > > > > > > I haven't looked at the details of what you are suggesting, but > > > > any change that slows 'quilt pop' down by a factor of 3 is not > > > > acceptable (to me). > > > > > > It would be nice to loose less speed but from the point of view > > > of a distribution maker, having it as a shell script simplifies > > > the bootstrapping of a new port: > > > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=363659 > > > > So, writing quilt in C would actually solve your problems :) > > > > Hm, if I ever get bored, that might be an interesting exercise...
I was thinking the same... Without that, we will be discussing performance issues without a reference point. But the problem is, I am not bored. At all. > we all know that having generic implementation is good thing. Linux > does have that, too. and having generics, in particular, helped me to > port Linux to new arch. > > but slowing quilt down just for that is not acceptable, IMNSHO. > > And, why don't we just do compile time check and install shell version > of backup-files if we do not support _that_ arch? does debian have to > have the shell version in all arch? That's one possibility, but it has drawbacks: we must maintain two versions of backup-files, we have to keep them fully compatible (which may prevent us from doing clean-ups that would speed up the shell counterpart), and we still have the bloated configure script. And the test suite would only test one variant at a time. So it doesn't sound terribly appealing to me. But I can be convinced, especially if the Debian folks say it would make their life much easier. -- Jean Delvare Suse L3 _______________________________________________ Quilt-dev mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/quilt-dev
