[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#471805: openoffice.org-core: spurious conflicts against older openoffice.org components



On 2008-03-20 11:11 +0100, Rene Engelhard wrote:

> Sven Joachim wrote:
>> In the last few updates of openoffice.org I noticed that
>> openoffice.org-core was always temporarily removed, it seems because it
>
> Because I don't upload i386 and the i386 buildd sometimes takes time...

I don't quite understand that.  Both -core and -calc are architecture
dependent, after all.  And all packages were upgraded in that aptitude
run.

>> unnecessarily conflicts with components from older packages, e.g.:
>
> Wrong.
>
>> openoffice.org-core Conflicts: openoffice.org-calc (<< 1:2.4.0~rc6-1)
>> openoffice.org-calc Depends:   openoffice.org-core (= 1:2.4.0~rc6-1)
>> 
>> This cannot be right, IMHO.  I've attached an excerpt from dpkg's log of
>
> Why? We go safe and do this to get the core and the apps using the core
> (and common) to have the exact same *upstream* version. Just that any
> last upload was a new version (rc3, rc4, rc5, rc6, ..).
> And we force the binaries to be the exact same version, too, but that
> doesn't affect your scenario here[1]
>
> This was introduced in 1:2.4.0~rc1-1. See the changelog:
>
>   * debian/control.in:
>     - fix logic error, we of course should conflict against old
>       openoffice.org-calcs in the new common, not the other way around
>       (really closes: #464544).
>       Go safe for the future; make -common Conflicts: against all the modules
>       (<< ${base-version})
>
> Read the bug to see what too lax depends can cause....

I've read it, but it does not really seem related, it is about -common,
not -core.  And the removals of -core predate this change, just run

zgrep 'remove openoffice.org-core' /var/log/dpkg*

to see that.

>> the latest upgrade which shows that it removed and reinstalled
>> openoffice.org-core, rather than upgrading it.
>
> Which is normal modus operandi of apt. Nothing to worry about.

I certainly won't lose sleep over that, but is generally safer to upgrade
packages rather than remove and reinstall them.  In the former case dpkg
can roll back the upgrade in case something goes wrong during unpacking,
in the latter you are left with a broken package.

> [1] Because if you have the new -core, you also have the new -calc, if
> not, you have the old versions there.

This is perfectly achieved by the versioned depends -calc has on -core.
The additional conflicts of -core on -calc is redundant.

See also bug #409411 for a similar instance in other packages.

Regards,
        Sven




Reply to: