[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [p2-dev] Question about generated ArtifactRepository

Hi Matthew,
I've done some debugging. While it's true that the source descriptor contains the artifact.reference appointing the .pack.gz, the destination computed from:

    private URI getLocationForPackedButFlatArtifacts(IArtifactDescriptor descriptor) {
        IArtifactKey key = descriptor.getArtifactKey();
        return mapper.map(location, key.getClassifier(), key.getId(), key.getVersion().toString(), descriptor.getProperty(IArtifactDescriptor.FORMAT));
    }

does not. Not when I copy a feature since there's no pack rule for it in the mapper. So the problem is even worse, I actually copy a .pack.gz into a .jar *without* unpacking it!

I haven't noticed this earlier since it got overwritten, but that's what happens. I'll add a bugzilla.

- thomas




Matthew Piggott wrote:

Hi Thomas,

The UpdateSite cheats a little with file locations, if you take a look at the generated artifacts.xml on each artifact there should be a property "artifact.reference" with the file location.

That said, there does appear to be something odd occurring when mirroring from an UpdateSite. Running a test locally it appears that packed files obtained from an UpdateSite are sent into the blobstore which seems a little strange. I'll take a look to see if I can come up with something.

Matthew

Inactive hide details for Thomas Hallgren ---04/23/2009 03:43:39 PM---Hi,Thomas Hallgren ---04/23/2009 03:43:39 PM---Hi,


From:

Thomas Hallgren <thomas@xxxxxxx>

To:

P2 developer discussions <p2-dev@xxxxxxxxxxx>

Date:

04/23/2009 03:43 PM

Subject:

[p2-dev] Question about generated ArtifactRepository





Hi,
I see some weird behavior when testing my mirroring approach of copying
the .jar.pack.gz files and regenerating the .jar locally. I debugged and
nailed it down to this:

1. My mirroring method finds two IArtifactDescriptors. One with no
format and one that is 'packed'.
2. I issue a RawMirroringRequest to copy the packed descriptor.
3. The unpacked descriptor is copied!
4. Restoring the .jar now results in an error. "Artifact repository out
of sync, overwriting ..."

As it turns out, this is because my source repository is generated from
a site.xml and it contains these three rules (and no more):

(& (classifier=osgi.bundle)), ${repoUrl}/plugins/${id}_${version}.jar
(& (classifier=binary)), ${repoUrl}/binary/${id}_${version}
(& (classifier=org.eclipse.update.feature)),
${repoUrl}/features/${id}_${version}.jar

I'm missing this one normally present in repositories created with the
publisher:

(& (classifier=osgi.bundle) (format=packed)),
${repoUrl}/plugins/${id}_${version}.jar.pack.gz

Well, in fact, I'm missing this one too:

(& (classifier=org.eclipse.update.feature) (format=packed)),
${repoUrl}/plugins/${id}_${version}.jar.pack.gz

or is it assumed that features are never packed? It wouldn't be much
gain since they normally don't contain any Java binaries but tools like
the site optimizer will pack features anyway I think.

My problem is that I have no way of detecting that this will happen.
I.e. there's no way of knowing that both ArtifactDescriptors represent
the same physical artifact since the rules are incomplete. Or is there?
If so, how do I do that?

TIA
- thomas
_______________________________________________
p2-dev mailing list
p2-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/p2-dev



_______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev

GIF image

GIF image