Nick Schurch wrote:
Does anyone have any idea why galaxy would pass a 'working directory' to the script rather than the directory where the files are actually stored? Is it that galaxy actually creates the files in the working directory and then moves them?
Hi Nick, This was originally written for PBS file staging - since the database/files/ directory was not mounted on cluster nodes, a path common to the Galaxy application server and the node's tool execution environment had to be used (the job working directory, which we put in NFS). The current method to use this functionality is such: If reading from an input's extra files, use $input.extra_files_path. The extra_files_path attribute always points to the location in database/files/. If writing to an output's extra files, use $output.files_path. The files_path attribute points to a directory in the job's working directory. The framework will automatically relocate these extra files to the real location in database/files/ as part of the job finish method. This functionality was never well documented and is rather confusing, in the future we plan to simplify it. A ticket exists for this: http://bitbucket.org/galaxy/galaxy-central/issue/206/eliminate-the-use-of-fi...
I'm also really confused as to why the 'development' installation works fine, but the 'live' installation throws this error. Any insight would be greatly appreciated.
This shouldn't really be the case, it might help to see your tool's xml and the code that generates the html, to understand how you're using the extra files directory. Note that links to files in the extra files directory should all be relative. --nate