On Sun, 21 Jun 2009, Matthias Kilian wrote: > And I managed to confuse my self a little bit...
lol > All I want to fix is the famous problem of bad checksums when > updating the shared-mime-info package. This happens whenever another > package like thunar-thumbnailers contains its own file in > share/mime/packages that overrides or extends mime-info entries > already in share/mime/packages/freedesktop.org.xml. Yes, I'm aware of the situation. Finding a reasonnable solution would be nice indeed. > Try this: install shared-mime-info, then install thunar-thumbnailers, > then update shared-mime-info (with pkg_add -r -Fupdate,installed). > > Some people suggested to `...@comment no checksum' entries like > share/mime/image/x-canon-crw.xml in shared-mime-info. espie@ objected, I was the one suggesting that. > because the mime-info stuff is too sensitive to be fiddled with by > other packages. I object, too, but for other reasons: > > The only relevant information is stored in share/mime/packages, > everything else is rewritten with each package addition, deletion > and/or update (for packages using @exec/@unexec update-mime-database), > so IMHO the problem is *not* a broken update-mime-database (as > landry@ said some days ago, IIRC), but with the files generated > from packages/freedesktop.org.xml included in the shared-mime-info > package. So you mean that all these other xml files will be created when packages @exec update-mime-database, right? > My diff changes that. And for the other ports, I just realized that > they already do the right thing: they include some > share/mime/frobnicate.xml and an @exec/@unexec update-mime-database > pair: > > $ find /usr/ports -path \*/pkg/P\* ! -path \*/shared-mime-info/\* | > xargs grep ^share/mime/. > > So, other ports don't need changes, except maybe minimal cleanup > of the plist (removal of share/mime/ and share/mime/packages/). Good, because that's what confused me in your last mail, I didn't see why any other packages would need to be modified. So, if I understand correctly, what you're proposing is pretty much the same as having @no checksum but in a more complicated way? Cheers! -- Antoine
