So atm. it's possible for create repo to create metdata which has the same nevra pkg (with the same pkgId) multiple times. Eg.
mkdir tmp mkdir tmp/x tmp/y for i in x y; cp pkg.rpm tmp/$i createrepo tmp ...this is specifically causing problems with excludes, as parsePackages() uses misc.unique on it's return values, which uses pkg.__hash__ which just uses repoid and envra. This means we only exclude one of the packages, and so then see the other one later. So here's my list of "obvious" possible solutions: 1. Add an argument to parsePackages() to give "unique" results based on just repo.id and pkg.pkgKey, used just be excludes atm. 2. Change createrepo to unique based on pkgId. 3. Change YumSqlitePackageSack.delPackage() to be based on pkgId (but still be a set). 4. Change YumSqlitePackageSack.delPackage() to be based on pkg.__hash__. 5. Change pkg.__hash__ to use repo.id and pkg.pkgKey, and deal with the fallout. ...now I think #5 is a bad idea (read a lot of work). It's possible that people can have two packages in the same repo. with the same envra data that have different pkgKey's and pkgId's (due to signing). So I don't think _just_ #2 fixes the problem, or #3. #4 might work, but means we need a pkg object -- or at least full repo/envra. I've done #1, and it's available: http://people.redhat.com/jantill/yum/yum-exclude-all-pkgs.patch ...and I think it's also worth doing #2 as well, because it just seems like a waste to have repeated data in the MD. Anyone else have any ideas? Or any different preferences to #1 and #2? [1] Can anyone think of a reason you'd have different pkgId/pkgKey but same envra and want to distinguish? -- James Antill <[EMAIL PROTECTED]> Red Hat
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Yum-devel mailing list [email protected] https://lists.dulug.duke.edu/mailman/listinfo/yum-devel
