On Mon, Aug 10, 2009 at 12:11:41PM -0400, Joey Hess wrote: > Wouter Verhelst wrote: > > When reading the documentation page, the '-a' option clearly states 'Act > > on all architecture dependent packages', and therefore this does seem to > > be consistent with the documentation. However, I believe redefining it > > to say 'all architecture dependent packages relevant for this > > architecture' would be more useful > > That's the -s flag, have you tried it?
Ah, no; I missed that. Thanks for pointing it out; I probably read the whole manpage at some point when I wrote the initial nbd rules file, but since that's so many years ago, I'm not surprised new features have cropped up in the mean time ;-) > > while it would be unlikely that this > > would bebreaking anything that depends on the current behaviour (the > > current behaviour results in either building the same thing as this > > proposed behaviour, or failing outright). > > IIRC the reason -a does not do what -s does is because it made debhelper > quite a bit slower since it needs to run dpkg-architecture (at least twice) > for each command. Oh, I see. Yes, that would be a good reason, I suppose. > That speed issue may not be very relevant now (except for > on, oh, m68k and arm?). And maybe that could be optimised.. in fact > I think I just eliminated it. > > The only case I can imagine where the current -a behavior might be depended > on is something using dh_listpackages -a to get a list of all binary packages, > including ones not built for the current architecture, and doing something > with it other than building packages. That seems unlikely. Right. > I don't think there's any particular reason for dh to have its current > behavior -- it can run dpkg-architecture just once when figuring out > which packages to act on. The change there looks to be this simple, though > I need to test it: ... and your follow-up mail shows that there is a significant difference. While detecing it could work, it might also be reasonable to just have dh support a '-s' option; if that is specified, then dh should pass -s rather than -a to all the commands it runs. OTOH, having a snippet that says override_dh_gencontrol: dh_gencontrol -s is not quite as ugly as the ifeq() make construction that I'm living with now, would not result in dh having to pass -s to *every* command so would not slow things down quite as considerably, and is small enough that I would be able to live with it. Feel free to close without further mention if you're not going to do the autodetection or extra option to dh. -- The biometric identification system at the gates of the CIA headquarters works because there's a guard with a large gun making sure no one is trying to fool the system. http://www.schneier.com/blog/archives/2009/01/biometrics.html -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org