On Fri, 2007-11-23 at 13:34 -0900, Jeff Spaleta wrote: > On Nov 23, 2007 12:31 PM, André Costa <blueser@xxxxxxxxx> wrote: > > Yeah, agreed about the resources. That's exactly why I am proposing > > that Fedora developers release early versions of anaconda's deps > > solver only -- not the whole thing. > > Isn't anaconda making use of yum's depsolver directly? My reading of > the bugzilla comments tells me this is reproducible with yum directly > without needing to use anaconda. > > So if depsolving problems are really yum bugs, don't we already have > enough tools being released as part of the testing process for people > interested in stress testing the depsolver without any extra releases > of anything? > > Comment 14 in the bug report has an example on how to yum reproduce > this specific issue. My gut feeling is that all these "26% upgrade hangers" stem from trying to upgrade an "FC7 Everything" (comprising FC7 updates) installation using the DVD image and yum encountering a bug during depsolving. Anyway, so far, I didn't encounter the "26% hanger" during my all FC7->FC8 upgrades I have performed so far (all on i386, only), but I noticed * upgrades corrupting the rpmdb (esp. due to broken %post scripts) * yum producing bizarre results on deps it could not resolve (esp. kernel related packages) * SELinux interfering into the upgrade process (Long series of complaints on invalid contexts and the like). I resorted to first selectively manually upgrade certain vital package (glibc, kernel, rpm, yum), then to reboot with selinux disabled and then to use "Everything+updates" instead of the DVDs. Ralf
