>Thanks for this advice. I eventually just made a list of the duplicates, >and used "rpm -ev" to remove them. I have been experimenting with >another package manager on my laptop, called smart[1], which works much >like yum, but has some improvements. It downloads repo metadata and >packages in parallel, making the process faster, but more interestingly, >as it upgrades packages, it removes the old one immediately. Thus, a >hung process would not leave the RPM database in such a mess. Perhaps >yum could be modified to work like smart, to guard against this kind of >problem. > Then I think smart is reordering the transaction set using its own ordering algorithm. To be clear the process that yum uses is: - pack the rpm ts - ts.order() - ts.run() so the order the packages are added/removed in transaction set is dependent on rpm, not yum. -sv
