On Wed, Apr 1, 2015 at 11:29 AM, Saket Choudhary <saketkc@gmail.com> wrote:
On 1 April 2015 at 02:06, Peter Cock <p.j.a.cock@googlemail.com> wrote:
On Tue, Mar 31, 2015 at 6:12 PM, John Chilton <jmchilton@gmail.com> wrote:
I think it is probably a pair of bugs you are seeing:
...
The fix unfortunately is messy and will likely break stuff.
-John
Running tools anyway where the expected example.loc entry is missing is bad, but most tools would fail cleanly when given an empty path - so this is less critical than the first half of the bug.
i.e. Can we address the side effects of Saket's commit?:
https://github.com/galaxyproject/galaxy/commit/667c04844e35e76a698161fff6c88... https://bitbucket.org/galaxy/galaxy-central/commits/757412c63654ff16f6cd6a0b...
Hi Peter,
Sorry, this really is a major bug. I agree this was pushed without much testing. In fact, I did not at all consider the use case you just pointed out. I can send a PR that simply reverts the change, since I am not sure I will have enough time to look into it deeper.
Looks like I have broken more things than what I have contributed.
Saket
Thanks Saket, I didn't mean to criticize - only to ensure you were aware of this (in case you had any insight about how to fix it). Galaxy is amazingly complicated, so subtle side-effects like this can be hard to avoid - especially on the interactive side where testing is harder. [We don't yet have a proper test framework for workflows - although John has been working on that.] I don't understand this area of Galaxy at all (mako templates etc), so I think we'll need to defer to the full time developers who know this bit best. Regards, Peter