Hide Forgot
+++ This bug was initially created as a clone of Bug #670215 +++ Description of problem: Rpm packaging of package 'report' forgot to specify _isa, which leads to mysterious errors when updating or installing other packages such as report-gtk. See https://bugzilla.redhat.com/show_bug.cgi?id=669842#c2, specifically: ----- 1. You have report-0.20-1.fc14.x86_64 installed. 2. You want to install report-gtk-0.20-0.fc15.x86_64, which has a dep. on repo = 0.20-0.fc15 (an older version than what you have installed). 3. Yum won't auto downgrade the x86_64 version, and the package doesn't use _isa, so yum thinks it can solve the problem by installing report-0.20-0.fc15.i686. ----- Version-Release number of selected component (if applicable): report-0.20-1.fc14.x86_64 How reproducible: every time Steps to Reproduce: 1. Attempt to install report-gtk.x86_64 0:0.20-0.fc15 when report-0.20-1.fc14.x86_64 is installed. 2. 3. Actual results: Error: Protected multilib versions: report-0.20-0.fc15.i686 != report-0.20-1.fc14.x86_64 Expected results: smooth integration between 'report' and 'report-gtk', including no complaints about i686 on an all-x86_64 system. Additional info: originally filed against yum as https://bugzilla.redhat.com/show_bug.cgi?id=669842
pushed upstream.
See Description on this bz; it has a description of what happens on x86_64 if you don't have _isa.
(In reply to comment #4) > See Description on this bz; it has a description of what happens on x86_64 if > you don't have _isa. Sorry, I wasn't thinking when I said this. You won't be able to test this (easily) more than you already have. This change will not have any significant effect until at least two versions of report go out with this change. At that point this change will prevent a customer from installing mismatched versions of (say) report and report-gtk. There has always been a version specific dependency of report-gtk on report, which should prevent a report/report-gtk version mismatch, but without this _isa change, yum would 'work around' the version specific dependency by installing a different version of report (one that matched the version needed by report-gtk). Of course, this version mismatch thing will never be a problem in the 'normal case' of a customer who only does updates from our official repos. It could only ever be a problem for a customer who try's to downgrade to an old version incorrectly, or has some glitch while yum updating, or installs some packages outside of yum's control, or if we make a mistake and put the wrong package versions into our repo, or the wrong packages versions get put on the satellight they are using ....
(In reply to comment #5) > Of course, this version mismatch thing will never be a problem in the 'normal > case' of a customer who only does updates from our official repos. "Never" is a very long time. The problem _will_ happen if there is more than one repository and they are at different over-all versions, or in a single repository if the packages are updated non-atomically. In practice is it hard to guarantee consistent versions due to reasons such as network delays, power outages, operator error or misunderstanding, etc.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2011-0703.html