On Tuesday 10 October 2006 18:05, Michael Biebl wrote: > reopen 391502 > thanks > > > On Friday 06 October 2006 18:45, Michael Biebl wrote: > >> Package: kdebase > >> Version: 4:3.5.5-1 > >> Severity: serious > >> Justification: violates policy > >> > >> The latest package update introduced a strict dependency on hal > >> and pmount. While the hal dependency is correct, the pmount > >> dependency is not. KDE simply calls the hal Mount()/Unmount() > >> methods, nowhere in the KDE code, pmount is actually used. If hal > >> would use pmount internally to do the actual mounting[1], fine, > >> let the hal package declare the correct dependencies. Adding the > >> pmount dependency to kdebase is simply wrong and violates the > >> policy, which states that strict dependencies have to reflect that > >> the "depended-on package is required for the depending package to > >> provide a significant amount of functionality" which is not the > >> case here. > >> Please see also #388341, the Recommends: pmount of > >> kdebase-kio-plugins does not directly violate the policy, still it > >> should be removed for the same reasons. > > > > If the user does not have a line in /etc/fstab for a drive, then > > HAL will not be able to mount the drive in a way that works > > properly with KDE. Strange but true. That is, of course, unless > > pmount is installed, in which case HAL uses it to mount devices, in > > a way which works with KDE. Thus the depends/recommends on pmount > > makes perfect sense, and is by no means a policy violation. > > Seems, you didn't read my email carefully. Since version 0.5.8, hal > does not use pmount anymore to do the mounting. Please check the > source code, if you don't believe me. > So now I have pmount installed, without any good reason. Argh, you're right, I didn't read the last blurb. Thanks. That said, pmount is still used (_directly_ by KDE) in some code, so while I'll drop it from kdebase's Depends, it'll stay as a Recommends with kdebase-kio-plugins. > And still, it's up to the hal package, to add the dependency on > pmount or not, because only the hal package knows, which tools it > uses internally to do the mounting. I still disagree here, because this was an unusual case where the backend's choice of tool affected the frontend in an important way. But now it doesn't matter anyway. Cheers, Christopher Martin
Attachment:
pgpePc9rnibmH.pgp
Description: PGP signature