Had a friend that accidentally upgraded to this version of rpm (rpm-2.91-17.i386.rpm) on a 5.x system and had lots of problems. if one does rpm -i --force package.rpm on a package that already exist,s it reinstalls the package, as it ought to, but makes another entry in the rpm database. This is a royal pain to fix, since you can't remove it normally ( you need to use --allmatches) and then resinstall the packages. Not sure if it's relevant or fixed already, just thought I'd mention it.
This won't happen if you use -U --force instead of -i --force.
Fine, but this behavior differs from rpm-2.5.3-5.1, and the behavior, though not completely irrational isn't rational either. Also, I suspect -U will remove packages of lower versions, which may not be desireable.
Ah, you need to do rpm --rebuilddb when switching from rpm-2.5.x to rpm-2.9x. In rpm-2.9x filenames are stored relatively, while in rpm-2.5.x filenames are absolute. At least I think that's what the problem is. Please reopen this bug if I'm wrong.
the --rebuilddb was run several times, in order to try to fix the problem. Let me reiterate that rpm -i --force package.rpm does "the right thing" in not producing duplicate entries, while 2.91-17 produced duplicate database entires. The duplicate entries were impossible to remove without ripping out the rpms and putting them back in again. I can't imagine why this is desireable behavior. The reason we were doing this in the first place was to install rpms that were of the same version but a "different build" Anyway, we "fixed" the problem by downgrading rpm, I just thought you'd like to know we found this bug, and it probably needs to be fixed eventually (though not for my benefit). If you still think it's not a bug then go ahead and close it.
ah shoot. > Let me reiterate that rpm -i --force package.rpm does the "right > thing" should read > Let me reiterate that rpm -i --force package.rpm under rpm-2.5.3-5.1 > does the "right thing"
I cannot reproduce this problem with rpm-2.93-2. I think the original problem was caused by either neglecting to run rpm --rebuilddb when upgrading to rpm-2.9x or bugs in rpm-2.91 Thanks for the report.