On Sun, 30 Nov 2003 06:56:50 +0100, Michael Schwendt wrote:
> On Sun, 30 Nov 2003 05:37:37 +0000, Keith G. Robertson-Turner wrote:
>> I just used apt to fetch pan, and it superseded 1:0.14.2.90-0.fdr.1 with
>> 1:0.14.2-1.rhfc1.dag.
>
> I doubt that.
Nope, I double checked and my package (on the website) definitely has an
Epoch of 1, and apt rejected it. If I comment out all the other repos,
then apt fetches it.
rpm -ev pan (ok)
rpm -ivh /var/cache/yum/base/packages/pan-0.14.2-2.i386.rpm (ok)
apt-get update (ok)
apt-get --dry-run install pan (pan is already the newest version.)
ls /var/cache/apt/archives/pan* (No such file or directory)
apt-cache showpkg pan
1:0.14.2.90-0.fdr.1 (mine)
1:0.14.2-2 (freshrpms)
1:0.14.2-1.rhfc1.dag (dag)
cat /etc/apt/sources.list | grep genesis-x
(rpm http://www.genesis-x.nildram.co.uk/apt fedora/1/i386 stable unstable testing)
md5sum temp/pan-0.14.2.90-0.fdr.1.i386.rpm
(f5d1a77226e2afb1d8b9336f4a072f60)
wget http://www.genesis-x.nildram.co.uk/apt/fedora/1/i386/testing/RPMS/
pan-0.14.2.90-0.fdr.1.i386.rpm (ok)
md5sum pan-0.14.2.90-0.fdr.1.i386.rpm
(f5d1a77226e2afb1d8b9336f4a072f60)
rpm -Uvh pan-0.14.2.90-0.fdr.1.i386.rpm (ok)
rpm -q --qf "%{EPOCH}:%{VERSION}-%{RELEASE}\n" pan (1:0.14.2.90-0.fdr.1)
So now apt thinks 1:0.14.2 supersedes 1:0.14.2.90
> No. The release is irrelevant in this case, since 0.14.2.90 is greater
> than 0.14.2 already. RPM versiong comparison splits %{version} at every
> dot and then compares the segments to eachother. The package version with
> the first higher segment wins. Since 0.14.2 and 0.14.2.90 are equal up to
> the last segment "90", the longer version wins. If both packages' versions
> are equal, RPM version comparison continues with comparing package
> %{release}.
>
> %{epoch} comparison overrides %{version}-%{release} comparison.
> Package with higher Epoch wins regardless of version-release.
That certainly makes sense, but as I said, it isn't happening.
Is apt broken?
