Back to the main page.
Bug 2062 - resolve some of the recent regression problems
| Status | CLOSED FIXED |
| Reported | 2013-03-17 14:10:00 +0100 |
| Modified | 2013-04-23 12:06:29 +0200 |
| Product: | FieldTrip |
| Component: | core |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Mac OS |
| Importance: | P3 normal |
| Assigned to: | Robert Oostenveld |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | |
| See also: |
Robert Oostenveld - 2013-03-17 14:10:48 +0100
the dashboard shows the following test script that recently started failing test_bug1129 test_bug1911 test_bug2059 test_example_preprocessing_eeg test_ft_topoplotER
Robert Oostenveld - 2013-03-17 15:08:56 +0100
regarding test_ft_topoplotER first problem is that the cfg is a config-object and that cfg.channel{:} fails (as Roemer has reported before) second problem is that non-unique channel names are present. Both have to do with my change in http://code.google.com/p/fieldtrip/source/detail?r=7672 http://bugzilla.fcdonders.nl/show_bug.cgi?id=2059
Robert Oostenveld - 2013-03-17 15:22:25 +0100
(In reply to comment #1) commit 0788559456f11cfd01cd9d64eace59df9831a078 Author: Robert Oostenveld <> Date: Sun Mar 17 15:16:24 2013 +0100 enhancement - improved channel handling in case of config object, see #2061 commit 68b4c3c4698c341174533e9511eb53ea634f0709 Author: Robert Oostenveld <> Date: Sun Mar 17 15:15:01 2013 +0100 enhancement - instead of giving an error, give a warning when non-unique channels are present and discard them. See also #2061
Robert Oostenveld - 2013-03-17 15:31:01 +0100
regarding test_bug1911 this also had to do with http://code.google.com/p/fieldtrip/source/detail?r=7672 http://bugzilla.fcdonders.nl/show_bug.cgi?id=2059 ans specifically with changing the variable name from lay to layout. The load(filename.mat) resulted in lay rather than layout. fixed in commit 6dbed3b6bbef8c357ba106831f32b772b0a5276d Author: Robert Oostenveld <> Date: Sun Mar 17 15:29:57 2013 +0100 bugfix - read lay variable from mat file and rename to layout, see #2061
Robert Oostenveld - 2013-03-17 15:33:31 +0100
test_bug2059 was simply due to not finding the mat file with the test data fixed. commit 5df4a0a5db1695474d646a2181b768403ecae014 Author: Robert Oostenveld <> Date: Sun Mar 17 15:32:51 2013 +0100 enhancement - read the test data from the correct location, see #2061
Robert Oostenveld - 2013-03-17 15:34:30 +0100
test_example_preprocessing_eeg works again, probably it also suffered from ft_prepare_layout
Robert Oostenveld - 2013-03-17 15:36:41 +0100
test_bug1129 has to do with filtering, which I have also been working on :-( Error using filter_with_correction (line 44) Calculated filter coefficients have poles on or outside the unit circle and will not be stable. Try a higher cutoff freque type/order of filter. Error in ft_preproc_bandpassfilter (line 128) filt = filter_with_correction(B,A,dat,dir); Error in test_bug1129 (line 217) y1 = ft_preproc_bandpassfilter(x1, 1000, [2 6]);
Robert Oostenveld - 2013-03-17 15:42:16 +0100
(In reply to comment #6) this is not due to the instabilityfix, but due to the change in the organization of the signal processing toolbox functions >> which butter /opt/matlab2012b/toolbox/signal/signal/butter.m >> addpath ~/matlab/fieldtrip-git/external/signal/ >> which butter /home/mrphys/roboos/matlab/fieldtrip-git/external/signal/butter.m With the mathworks signal/butter I get the error, with the replacement butter I don't get it. So apparently the two implementations are not identical.
Robert Oostenveld - 2013-03-17 16:00:15 +0100
(In reply to comment #7) See http://bugzilla.fcdonders.nl/show_bug.cgi?id=2033 http://bugzilla.fcdonders.nl/show_bug.cgi?id=283 I compared the filter parameters: A is slightly different, causing the poles>1. Apparently the octave version is slightly better than the mathworks version The solution is to always use external/signal. This also saves some licenses at the DCCN. fixed. commit fcba8a0fa95fa2ade0e193123ba89da0add47e7c Author: Robert Oostenveld <> Date: Sun Mar 17 15:58:38 2013 +0100 enhancement - always add external/signal, see #2033 and #2061
Robert Oostenveld - 2013-03-17 16:06:53 +0100
wrapping up: http://code.google.com/p/fieldtrip/source/detail?r=7683 contains all the commits that I made in git and merged into svn. That was not what I had in mind, as you cannot see the details any more. Well, hopefully the dashboard shows some improvements.
Robert Oostenveld - 2013-04-23 12:05:53 +0200
closed various bugs