Charles A Edwards <[EMAIL PROTECTED]> writes: > On Fri, 10 Oct 2003 17:06:25 +0200 > Jan Ciger <[EMAIL PROTECTED]> wrote: > > > 3) the behavior of RPM in this case is strange (it is an RPM not urpmi > > issue). > > It is not strange. > > A looked at the posted syslog provides the answer. > > At the time the upgrade was done there existed a flux in cooker tree > which required that pam-0.77-9mdk be removed. > When pam-0.77-9mdk is removed it causes system-auth to > to be changed to rpmsave. > When then pam-0.77-10mdk is installed it will create > system-auth.rpmnew, believing that system-auth still exist. > > The stated occurrence in an aberration and could only occur under very > limited and specialized circumstances. > > Within the past hr I used --auto-select to install 198mb of updates and > experienced no problems in the pam upgrade. > But there was a problem with libtool, as libtool-1.4.3-8mdk was present > in the hdlist but neither libltdl3-1.4.3-8mdk nor > libltdl3-devel-1.4.3-8mdk were.
Thanks Charles ! It will help me tracking down this bug of rpm (and not urpmi for sure), there are various way to fix it, workaround in pam itself (make sure a system-auth file is present), workaround in urpmi (in the same way problably), real fix in rpm (which is bar far the best solution but problably the hardest :-(). There is finally a last try, if somebody can try it, a bad invocation of urpmi on the package, this should be available using -v flag which will display what urpmi is putting in transaction that is causing rpm to behave strangely, I will try it monday if nobody can test it. But I suspect nothing is bad on this side, at least I hope for... Francois.
