Re: [p2-dev] Galileo mirroring problem. Need help!

[Thomas Hallgren <thomas@xxxxxxx>]

David M Williams wrote:

The file was reconditioned with a Java 1.5 Pack200, and then signed at 
    

Eclipse.org, and then packed again with a Java 1.5 Pack200.

That is a little different than the flow many of us use. Since the 
Eclipse.org script also does a re-pack (conditioning) many of us do not do 
that before submitting to be signed. While this shouldn't be a problem (to 
-repack twice) it might expose some other bug the rest of us do not see. 
  

Hmm, if the signing is Java 6 at Eclipse.org now, could it be that the repack is Java 6 too?

BUT, you say you can 'manually' unpack and verify the pack.gz file, so I 
think it must be more than that. 

  

Any ideas are greatly appreciated. I'm starting to drain out of them.
    

Ok, you asked, so if it is any ideas you want, here's some more wild ones. 
I hate to suggest this to _you_ ... but are you sure the files the build 
is "getting" are really in the same repo as the ones you are checking 
(i.e. urls in artifacts.xml/jar are correct?

The builder mirrors the files into the buildresult/final/aggregate repository. I made my checks right there.  I also renamed the last buildresult to buildresult.old so that it doesn't go away in case anyone else would like to try the same.

  Also ... since this seems to 
happen for buckminser, and buckminster is currently installed in the 
Platform that is running the build, any chance that is causing some 
mix-up?

I thought about that too. But buckminster is not the only thing installed. The Galileo builder feature also contains a bundles from EMF. I'm also quite certain that the artifact being unpacked is the artifact that was just mirrored since the mirroring logic deals with one single repository at a time.

- thomas