Back to the main page.

Bug 2835 - bug binge: diagnose and/or fix broken testscripts that were still working recently

Reported 2015-02-11 11:15:00 +0100
Modified 2019-08-10 12:32:53 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P1 normal
Assigned to:
Depends on:
See also:

Robert Oostenveld - 2015-02-11 11:15:07 +0100

I identified the following test scripts that fail with r10199 but that somewhere between r10159 and r10198 were still working. That means that a recent change was introduced that breaks the scripts. The change to the code should be pinpointed, as it is likely that with the change fresh in mind it will be easier to fix it. test_bug1646.m test_bug1820.m test_bug1856.m test_bug1954.m test_bug2336.m test_bug2338.m test_bug2397.m test_ft_componentanalysis.m test_ft_prepare_mesh.m test_headmodel_simbio.m

Eelke Spaak - 2015-02-11 14:13:44 +0100

Assigning this bug to general user again, all the scripts listed in the original bug report have to do with code about which I know very little. I will report other test scripts that I worked on here.

Eelke Spaak - 2015-02-11 16:22:19 +0100

test_tutorial_spike_Neurosim: fixed test_bug1708: data not found (see bug 1708) test_bug1742: dccn-specific cluster issue, workaround by backend='local', is fixed now test_bug1794: test script is weird (see bug 1794) test_bug2096: permission denied when running it (not the error on the dashboard) test_bug2185: priority bug in getdimord, fixed now bash-4.1$ svn commit private/getdimord.m test/test_tutorial_spike_Neurosim.m test/test_bug1742.m Sending private/getdimord.m Sending test/test_bug1742.m Sending test/test_tutorial_spike_Neurosim.m Transmitting file data ... Committed revision 10208.

Jim Herring - 2015-02-11 16:29:46 +0100

I've looked into test_bug1646.m and test_ft_prepare_mesh. Something seems to go wrong when cfg.tissue is indexed instead of having character labels. At some point the tissue labels 'tissue1', 'tissue2', and 'tissue3' are created, which are then removed by convert_segmentationstyle and replaced with 'inside'. This then crashes prepare_mesh_segmentation because it expects as many labels as there are tissue indices.

Jim Herring - 2015-02-11 16:52:57 +0100

I've narrowed it down to ft_checkdata adding a field segmentation.inside, which is detected as a probabalistic field later on. In prepare_mesh_segmentation, the following line tries to convert this probabalistic .inside field into an indexed field while replacing the tissue labels that were in place for the actual indexed field. 98 mri = ft_datatype_segmentation(mri, 'segmentationstyle', 'indexed');

Eelke Spaak - 2015-02-11 16:58:03 +0100

test_bug2355: .inside numeric -> logical still working on bug 2381, but have to go now bash-4.1$ svn commit test/test_bug2355.m Sending test/test_bug2355.m Transmitting file data . Committed revision 10210.

Robert Oostenveld - 2015-02-11 17:47:37 +0100

test_bug1646.m is now working

Robert Oostenveld - 2016-02-05 13:13:18 +0100

this was for a specific binge, and can be closed

Robert Oostenveld - 2019-08-10 12:32:53 +0200

This closes a whole series of bugs that have been resolved (either FIXED/WONTFIX/INVALID) for quite some time. If you disagree, please file a new issue on