Trusted, Bureaucrats
4,228
edits
Changes
→Bugfixes: Copied the bugfix list from the CHANGES file.
<section begin=bugfixes/>
* Replicated titles in the HTML version of the relax manual (http://www.nmr-relax.com/manual/index.html), and hence replicated HTML file names overwriting earlier sections, have been eliminated.
* Fix for bug #23933 (https://gna.org/bugs/?23933). This is the "NameError: global name 'ids' is not defined" problem when loading relaxation data. The bug was introduced back in November 2014, and is due to some incomplete error handling code. The problem is that the spin type that the relaxation data belongs to (@N vs. @H) has not been specified. Now the correct RelaxMultiSpinIDError is raised. The 'ids' variable did not exist - it was code that was planned to be added, but never was and was forgotten.
* Fix for the CSA constant equation in the model-free chapter of the manual. This was spotted by Christina Möller and reported on the relax-users mailing list at https://mail-archive.com/relax-users%40gna.org/msg01776.html .
* Bug fix for the storage of the XML structural object in the state and results files. Previously any objects added to cdp.structure (or any structure object) would not be saved by the structural object to_xml() method unless the function is explicitly modified to store that object. Now all objects present will be converted to XML.
* Fix for the relaxation dispersion analysis in the GUI, as caught by the Relax_disp.test_paul_schanda_nov_2015 GUI test. When loading from a script state file, the value of "None" can be present. This is now set to the standard values.
* Fix for running relax at a server with no graphical display and using matplotlib. The error was found with the Relax_disp.test_repeat_cpmg system test. And the error generated was: "QXcbConnection: Could not connect to display. Aborted (core dumped)". The backend of matplotlib has to be changed. This is for example described in: http://stackoverflow.com/questions/2766149/possible-to-use-pyplot-without-display http://stackoverflow.com/questions/8257385/automatic-detection-of-display-availability-with-matplotlib.
* Modified the behaviour of the bmrb.write user function backend for a model-free analysis (fix for bug #24131, https://gna.org/bugs/?24131). This is in the bmrb_write() method of the model-free analysis API. Deselected spins are now skipped and a check has been added to be sure that spin data has been assembled.
* Another fix for bug #24131 (https://gna.org/bugs/?24131). This is the BMRB export failure when the SpinContainer object has no s2 attribute. Now no data is stored in the BMRB file if a model-free model has not been set up for the spin. This allows the test suite to pass.
* Bug fix to allow the Mf.test_bug_24131_missing_interaction system test to pass. This is part of bug #24131 (https://gna.org/bugs/?24131), the BMRB export failure with the SpinContainer object having no s2 value. The problem was when assembling the diffusion tensor data. The spin_loop() function was being called, as the diffusion tensor is reported for all residues. Therefore the skip_desel=True has been added to match the model-free part.
<section end=bugfixes/>