The copy pattern is **/*

Something does indeed look amiss.

I think I've seen this problem before for some temporary files. i.e. the file being moved, were actually removed by another process while Bamboo was trying to copy it.

Your setup seems fine to me. A few questions

1. Does this problem always occur? Or is it intermittent?
2. Can you make sure that the user running Bamboo has read access o /home/bamboo-server/home/xml-data/build-dir/APP-MAIN/output/db/classes/com/networks/db/DatabaseRuntimeException.class and write access to {[/home/bamboo-server/home/xml-data/build-dir/APP-MAIN/output/db/classes/com/networks/db/DatabaseRuntimeException.class}}. I wonder if it may be a permissions thing.

Cheers,

Mark C

2. Everything for the server runs as "bamboo-server".

This user has r/w access to the source file in question. They also have write access to the target directory and can write a file to this directory without any problems.

I don't think the class files are temporary - they still exist after the server has finished working and after it has complained about them not existing.

I'll try restarting the server with an strace and see if I can see where the server is looking.

Interestingly, I discovered this in the log this morning:

14-Dec-2006 20:15:02 Executing build 108
14-Dec-2006 20:15:02
Starting to build 'project - Main Build'
... running command line: /opt/apache-ant-1.6.5/bin/ant -f build.xml clean compile compile-tests comprehension
... in : /home/bamboo-server/home/xml-data/build-dir/PROJECT-MAIN
... using java: /usr/java/jdk1.5.0_10
... using environment variables:
14-Dec-2006 20:15:03 Buildfile: build.xml
14-Dec-2006 20:15:14
14-Dec-2006 20:15:14 clean:
14-Dec-2006 20:15:14 [delete] Deleting directory /home/​bamboo-server/​home/​xml-data/​build-dir/​PROJECT-MAIN/​output
14-Dec-2006 20:15:14 Failed to copy build artifacts using pattern "*/"

Now why is it trying to do the artifact copy at the START of the build process?

Seems very very weird, some kind of parallelism problem perhaps?

Looking at the code again, it does indeed look like a possible multi-threaded issue. Currently, we deal with artifacts as a build post process, which means that it's possible that another build could be kicked off before the post processes would occur. This wouldn't be such a problem with notifications and the like, but obviously is a problem with artifacts (since the files are being ripped from underneath it).

Thanks for helping us out on this. I'll see if I can patch it for 0.9

Cheers,

Mark C