Peter, As it turns out, the automated testing framework was not correctly handling the situation where dataset metadata is set externally. I've committed a fix for that in 10295:59243394031a, and this should result in seq_rename now being correctly tested. When it comes to the sniffer situation, I've created a trello card for that issue, where you can track the status: https://trello.com/c/eCRFKVkt/1024-when-a-file-is-uploaded-upload-py-runs-it... --Dave B. On 8/2/13 05:14:54.000, Peter Cock wrote:
On Thu, Aug 1, 2013 at 8:43 PM, Dave Bouvier <dave@bx.psu.edu> wrote:
Peter,
As of 10290:d34d6d1da813, nsltradamus and effectivet3 should start returning valid test results.
Great - so this was a sniffer bug for the twobit format going wrong, https://bitbucket.org/galaxy/galaxy-central/commits/d34d6d1da813bfa6e318b0ce...
As for seq_rename, it looks like the columns in four_human_proteins.rename.tabular are not being detected correctly, even though they are tab separated. I'll continue to investigate, and let you know what I discover.
--Dave B.
I was assuming (wrongly perhaps) that if the tool XML for the test defined an ftype this would be used as the datatype - overriding any sniffer guess (or even skipping the sniffer tests).
(You can think of this a like testing the process where a user has uploaded a given file and manually overridden the datatype.)
i.e. It is good that this issue has apparently uncovered some sniffer bugs, but that doesn't seem to be the root issue.
Thanks,
Peter