On Monday 24 November 2003 14:27, seth vidal wrote: > you didn't respond to my other question - what if it is not a yum update > but a yum update foo* bar baz Nothing special, imho. If fooA and fooB are installed, then foo* is expanded into 'fooA fooB' before anything else is done, right? Now, if fooB and bar cannot be updated, due to dependency problems, but fooA and baz can, why not upgrade these? The dependency problems for fooB and bar are in no way fatal for fooA and baz. > > Installing fewer packages than requested should not be too shocking, > > given that all other dependencies are satisfied. Yum should output a > > warning, though. > > Why isn't it shocking? If I've just asked my computer to do X and it > does X-3 I'm not surprised? No. If I ask my computer to do X, Y, Z, but X and Z won't work, what do I expect? It depends: if I meant 'X && Y && Z', the whole command should fail, but if I meant 'X; Y; Z', then Y will do fine. In the case of 'yum update a b', I (personally) would expect yum to do (sort of) 'yum update a; yum update b' and not 'yum update a && yum update b' > And should it sys.exit(0) or sys.exit(1)? It seems like an error state > to me. why should it complete successfully? Ok. If 'update' reads 'update-all' then it is erroneous not to install all requested packages, I must admit. On the other hand, there is already the switch '-t'. In tolerant mode, yum complains but doesn't fail on update requests that can not be fulfilled (but this is limited to errors on the cmdline). Maybe the behaviour I propose should be restricted to tolerant mode. Thomas
