Hi,
It seems the cuffcompare XML wrapper expects a file named "cc_output" to contain the "transcript accuracy" report. I guess it worked with versions 0.9.3 and 1.0.1, but at least in version 1.0.2, this file is now named "cc_output.stats", and the "transcript accuracy" file is empty.
so the XML wrapper is compatible with 1.0.1 but not really with 1.0.2/3 (Unless I missed something in the XML/python/cuffcompare execution chain).
Also, Notice that starting version 1.0.2 they've added an option to normalize by either total number of reads or by number of reads that matched the given GTF file. To make matters worse, "cufflinks" defaults to normalizing to total reads, and "cuffdiff" defaults to normalizing by GTF-matching reads. The options are called "--total-hits-norm" and "--compatible-hits-norm".
There's no way for users to specify either of those in Galaxy, so running cufflinks/cuffdiff might give unexpected results (not that there are any "expected" results, but still...)
Just something to think about when you work on the next version of cufflinks wrappers.
-gordon