[Rpm-maint] rpm --verify and replaced files

Panu Matilainen pmatilai at laiskiainen.org
Tue Apr 9 11:04:38 UTC 2013


On 04/09/2013 11:40 AM, Michael Schroeder wrote:
> On Tue, Apr 09, 2013 at 08:32:57AM +0300, Panu Matilainen wrote:
>> I suppose the idea has been not to complain when its "known broken", ie
>> installed with --replacefiles or --force. The kinda legitimate use-case for
>> that would be replacing %config files from another package, for everything
>> else forced files pretty much of course equals broken package. I agree it
>> would be good to somehow note this in the --verify output, but since the
>> verification output is the way it is... ideas welcome.
>>
>> One possibility might be adding another "field" to the output where any
>> non-regular state is shown as a one-letter abbreviation (such as R for
>> replaced, W for wrongcolor etc).
>
> Yes, that would work for me. Another easy bandaid is to add an option
> to make verify report bad replaced files.

I'd prefer having rpm do something reasonably sane without even more 
switches :)

There are some other peculiar cases wrt verify too: if a file was 
intentionally not installed (eg --nodocs or --noconfigs), currently 
verification just completely ignores the file, returning "everything 
ok". But what if such a file actually exists on the filesystem? Seems to 
me verify should actually complain about it. Also whereas flagging 
"replaced" state in verify seems fairly obvious, its not entirely clear 
whether "known missing" should be flagged there as well.

Another mystery is what to do with RPMFILE_STATE_NETSHARED: rpm could 
try verifying them as such files are actually expected to be present, 
just not managed by this rpm instance. But then you wouldn't want 5.000 
clients verifying (digest calculation and all) the same files over a 
shared NFS mount.

	- Panu -



More information about the Rpm-maint mailing list