Hi,
I add a look and something puzzle me here.

I cannot reproduce the problem. Note I have added unit test in this commit [1]
Did I miss something with your use case ?

Thanks
-- 
Olivier Lamy
http://twitter.com/olamy | http://linkedin.com/in/olamy

[1] http://svn.apache.org/viewvc?view=revision&revision=1325558

2012/4/10 Sascha Vogt <[email protected]>:
> Am 06.04.2012 11:35, schrieb Olivier Lamy:
>> 2012/4/6 Sascha Vogt <[email protected]>:
>>> Hi,
>>>
>>> Am 05.04.12 11:46, schrieb Olivier Lamy:
>>>> 2012/4/5 Sascha Vogt <[email protected]>:
>>>>> It seems that the purge deleted the snapshot -javadoc.jar but left the
>>>>> -javadoc.zip and we ended up with several thousand files in one
>>>>> directory which slowed down the repository scan to take ages.
>>>>>
>>>>> The question is: Can or should Archiva either
>>>>> a) refuse deployment if the same classifier is deployed twice with a
>>>>> different extension
>>>>> or
>>>>> b) the purge be enhanced to delete all extensions of a given classifier?
>>>> I tend to say b) :-)
>>>> Can you load an issue for this feature ?
>>>> BTW is **/*.zip recorded as a file pattern in your archiva configuration ?
>>> created http://jira.codehaus.org/browse/MRM-1622 and yes **/*.zip is
>>> added as a file pattern.
>>>
>>> I'll try to have a look next week and provide a patch.
>>
>> Cool let us know if any questions :-)
> I do ;)
>
> I tracked the issue down a little bit further. It seems to be an issue
> relating to the type <-> extension mapping.
>
> The purge consumer gets a list of related artifacts
> (repository.getRelatedArtifacts()) which contains ArtifactReferences to
> the files to be purged. An ArtifactReference has no file extension, but
> only a type. So if you use a "wrong" extension for a given type (e.g. as
> in our case zip for javadocs) the conversion back to an actual file will
> then produce a file, which is not existing...
>
> Frankly I have NO idea how to fix that with an easy patch :).

>Maybe extend the ArtifactReference with an optional extension and use that
> when converting between files and references if present?
>
> Or should the ManagedRepositoryContent be extended by a method which
> directly returns files instead of references (omitting the conversion
> from path to reference and back to a path)?



>
> Any input is welcome!
>
> Greetings
> -Sascha-

Reply via email to