Re: [egit-dev] galileo build broken

[David M Williams <david_williams@xxxxxxxxxx>]

> we should not
have Europa dependencies. Is there a simple way to check this other than
checking every bundle manually ?
I do not know. grep for 'europa' in
any files that would reference a p2 url? I really was "grasping at
straws", just in case it reminded you of anything ... unlikely to
be a factor. 

> This also fails
quite often since a few days, rerunning 
> the egit build usually yields a successful
build.

Even for intermittent bugs, you should
open a bug on Eclipse Foundation (servers component, or, I guess hudson,
if you think related to "slaves" or what not) with as much info
as possible, including times, URLs involved, preferably shortly after it
happens so info in error or server logs (for example) might be easier to
spot. ... occasionally there are glitches (that I do not completely understand)
where part of the "internal" network between Eclipse Foundation
nodes will be broken, but not all legs of the network, so it is a matter
of probability if your request is routed through a good or bad path. But,
honestly ... I'm just guessing and suggesting the little I know. It could
be completely something else. Hard to debug. Sorry I can't help further.





From:      
 Matthias Sohn <matthias.sohn@xxxxxxxxxxxxxx>
To:      
 David M Williams/Raleigh/IBM@IBMUS
Cc:      
 EGit developer discussion
<egit-dev@xxxxxxxxxxx>
Date:      
 12/10/2010 03:01 AM
Subject:    
   Re: [egit-dev]
galileo build broken

---

2010/12/9 David M Williams <david_williams@xxxxxxxxxx>
[quoted from cross-project list] 
> Was there any infrastructure change on hudson.eclipse.org
which might cause such failure ?
> I found that only our build against galileo is affected, the one against
helios works.  
> The local build against same galileo p2 repository works. No clue
what's wrong here. 

We did delete a junk directory from /releases/europa ... but that shouldn't
effect galileo based builds. I just mention it, in case that info makes
something click for you ... like maybe it's really a Europa based build?
or you have some stray Europa dependency? (and ... still ... what we deleted
should not effect a Europa based build either ... just grasping at straws).


we should not have Europa dependencies. Is there a simple
way to check this other than checking every bundle manually ?

Perhaps a better train of thought ... Galileo was back
in the days we did not always store the 'jar' version of a bundle, but
sometimes only its pack.gz version: 

$ find ./ -name "*org.easymock*" 
./galileo/200909241140/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar.pack.gz
./galileo/201002260900/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar.pack.gz
./helios/201006230900/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar.pack.gz
./helios/201006230900/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar
./helios/201009240900/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar.pack.gz
./helios/201009240900/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar
./indigo/201008200901/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar.pack.gz
./indigo/201008200901/aggregate/plugins/org.easymock_2.4.0.v20090202-0900.jar

I just mention this, in case that provides a hint as to maybe something
else that changed (on your end?) that would interact with "jar"
vs. "jar.pack.gz"? 
And, not sure how you created your "local" version ... but maybe
it has a jarred version of the bundle? 

We use Maven 3 and Tycho 0.10, local build just means
running the same maven build on a local box against the same
p2 repositories given in the egit parent pom.xml [1] after
having built jgit as described in the Contributor Guide [2]. 
To prevent caching effects from hiding the problem I also
tried this from a fresh empty local maven repository, still the 
local build succeeds.

When running the egit build on hudson.eclipse.org
we consume jgit through its p2 repository created by the jgit hudson job 
on the same box hudson.eclipse.org
via its p2 URL on hudson [3]. This also fails quite often since a few days,
rerunning 
the egit build usually yields a successful build.
When running the egit build locally we refer to the jgit p2 repository
through
a local file: URL.
 
[1] http://egit.eclipse.org/w/?p=egit.git;a=blob;f=pom.xml;h=86cb6a7a063902608142394696967585021ca3c8;hb=HEAD
[2] http://wiki.eclipse.org/EGit/Contributor_Guide#Builds
[3] https://hudson.eclipse.org/hudson/job/jgit/lastSuccessfulBuild/artifact/org.eclipse.jgit.packaging/org.eclipse.jgit.updatesite/target/site/

-- 
Matthias