i'm basing this on the doc for "Editing comps.xml" at http://rhlinux.redhat.com/anaconda/comps.html. if you read the section "Generating the full comps file", the first bullet point refers to a tarball called "comps-8.0.tar.gz" you can download that *allegedly* contains the master version of the comps file. ok ... that tarball contains two relevant files: comps-milan.xml.in comps-milan.xml and running "make" generates the second file from the first one. so i'm assuming that that second file is considered to be the master comps file. more to the point, it does not contain package information, which appears to agree with the hint listed that you should "maintain your master comps file without this information". so far, so good. this suggests that, if one needs to start building a new comps file, this master version represents a good starting point, right? now, the second bullet point says to run the command: # getfullcomps.py comps.xml /path/to/tree arch which is where the trouble starts. note the reference to a file called "comps.xml". given that there is, in this tarball, no file called "comps.xml", which file is this referring to? weirdly, the online docs state that this file must *already* exist in your Red Hat installation tree in its proper location. but that's silly. given that the whole point of this exercise is to generate a new comps.xml file, it seems absurd to have, as a prerequisite, that you must already have a file by that name in the eventual destination location -- that is, in /fullpath/i386/RedHat/base/comps.xml. but, sure enough, it has to be there. if it isn't, the getfullcomps.py pgm blows up with a python traceback. so, in a nutshell, this is frustrating/confusing for a number of reasons: 1) which file is considered to be the "master version" of the comps file? is it the one that can be created from the contents of the tarball? 2) why does the getfullcomps.py program require a pre-existing comps.xml file as a pre-requisite for creating a new one? 3) why doesn't the same program at least exit gracefully if that file doesn't exist, rather than bursting into flame with a python traceback? rday
