[relax-announce] relax version 4.1.2.

[Edward d'A. <ed...@nm...>]

This is a minor feature and bugfix release.  It includes tooltip
improvements in the GUI for the user function windows and wizards, the
addition of the newly published primary reference for the frame order
analysis ( https://doi.org/10.1017/S0033583519000015 ), and improved
formatting for the bibliography and index of the relax manual.

There have also been improvements for the automated testing of relax
by Travis CI ( https://travis-ci.com/nmr-relax/relax ).  This includes
the naming of the build jobs, the execution of the software
verification tests, the installation of wxPython to enable GUI testing
and the running of the whole test suite, the reordering of the system
tests back before the unit tests to avoid hiding some nasty relaxation
dispersion bugs, a fix for matplotlib on Mac OS X so that the tests
will finally run on this OS, a new build job for the API
documentation, and a new build job for the Free Software Foundation
copyright validation script.

For the official, easy to navigate release notes, please see
http://wiki.nmr-relax.com/Relax_4.1.2 .

The new relax versions can be downloaded from
http://www.nmr-relax.com/download.html. If binary distributions are
not yet available for your platform and you manage to compile the
binary modules, please consider contributing these to the relax
project (described in section 3.6 of the relax manual,
http://www.nmr-relax.com/manual/relax_distribution_archives.html).

The full list of changes is:

Features:
    * relax GUI: Improved tooltips for the buttons of the user
function windows and wizards.  This follows from the mailing list
discussion at https://sourceforge.net/p/nmr-relax/mailman/message/36611489/.
    * User manual: Addition of the newly published frame order reference.
    * Formatting improvements for the user manual bibliography and
index sections.


Changes:
    * Development scripts: Improvements to the Python detection in the
Python module seeking script.
    * Release checklist document: Updated the text to better match the
new release process.
    * HTML manual: CSS fix for newer LaTeX2HTML versions.  The text
width in the HTML appears to now be fixed to a maximum width matching
the text dimensions in the PDF.  This looks bad together with the
wider images and code snippets.
    * System tests: Added two tests to catch bug #12
(https://sourceforge.net/p/nmr-relax/tickets/12/).  This is the
failure to catch the '#' character when setting the molecule name.
The tests are Structure.test_bug_12_hash_in_mol_name_via_arg and
Structure.test_bug_12_hash_in_mol_name_via_file.  These cover the two
ways a '#' character can enter a molecule name - via the file name or
via the set_mol_name argument.  Both the structure.read_pdb and
structure.read_xyz user functions are checked.
    * Test suite: 2 new system tests to catch the failure of reading
newer Bruker DC files.  The system tests are
Bruker.test_bug_13_T1_read_fail and Bruker.test_bug_13_T2_read_fail
and these catch bug #13
(https://sourceforge.net/p/nmr-relax/tickets/13/).
    * User function definitions: Clarifications for the bruker.read text.
    * User manual: Clean up of the bibliography entry titles.  Species
names are properly italicised with genus names capitalised, nuclear
isotopes are superscripted, R1rho, R2, etc. are properly subscripted,
the Perrin articles are translated into English, symbols are now
symbols, and unnecessary capitalisation has been removed from the
bibtex.
    * User manual: Standardisation of the frame order indexing.
    * User manual: Standardisation of the relaxation dispersion indexing.
    * Travis CI config: Attempt at installing wxPython for Ubuntu and
Python 2.7.  This would allow for the whole test suite to be run on
Travis CI on at least one OS.  The instructions come from the
stackoverflow response by dthor at
https://stackoverflow.com/questions/29290011/using-travis-ci-with-wxpython-tests.
    * FSF Copyright Validation script: The script now returns an exit status.
    * Travis CI config: Avoid updating Conda.  This seems to cause a
breakage in installing matplotlib.
    * Travis CI: matplotlib is now manually installed to allow for
older versions on Python 2.7.  The current pip default of 3.0.3 is
incompatible with Python 2.7.  It is not clear how the installation of
Conda (for wxPython support) caused the 3.0.3 version to be installed
instead of the 2.2.4 version.  So now the version is manually set in
the Travis CI script.
    * Travis CI config: Enable xvfb to allow for wxPython and testing
of the GUI.
    * Test suite: Restored the original test suite order to reveal
relaxation dispersion bugs.  The system tests should come first.  This
allows the maximum amount of code that might accidentally change
read-only variables to run prior to the unit tests, where such changes
are often subsequently picked up.
    * Test suite: The keyboard interrupt terminates the test suite once again.
    * FSF Copyright Validation script: The return status now starts at
0 to allow for early returns.
    * FSF Copyright Validation script: Support for saving and reading
the committer information.  This allows the committer information
(file name, committer name, and copyright years) from older
repositories to be saved and later read into the script.  In this
case, the old Subversion history has been read and the committer
information placed into the fsfcv.svn_committer_info.bz2 file (in the
devel_scripts/ directory).  This compressed file is now specified in
the fsfcv.conf.py configuration file.  The result is that the fsfcv
script can be run on the relax git repository without requiring a
checkout of the old SVN repository.
    * Travis CI config: Improvements to the comments and spacing.
    * API manual: Scons compilation via epydoc now fails if a warning
or error is found.  This manually parses the epydoc output to skip the
unavoidable wxPython warnings.  Any error or warning will now cause an
error to be raised.  This results in a non-zero return code from scons
to allow the api_manual_html target to be checked in scripts.
    * Travis CI config: Named all of the jobs.
    * Travis CI config: General clean up and execution of the software
verification tests.
    * API manual: Scons compilation via epydoc now fails if an import
error occurs.
    * Travis CI config: Alphabetical ordering of environmental
variables and required Python packages.
    * Travis CI config: Creation of an API documentation build job.
    * Travis CI config: Fix for the Mac OS X build.  This job passes,
but the test suite fails with the following traceback message when
trying to import matplotlib: "ImportError: Python is not installed as
a framework. The Mac OS X backend will not be able to function
correctly if Python is not installed as a framework. See the Python
documentation for more information on installing Python as a framework
on Mac OS X. Please either reinstall Python as a framework, or try one
of the other backends. If you are using (Ana)Conda please install
python.app and replace the use of 'python' with 'pythonw'. See
'Working with Matplotlib on OSX' in the Matplotlib FAQ for more
information.". The fix is simply to create
$HOME/.matplotlib/matplotlibrc with the contents 'backend: TkAgg'.
    * API manual: Greater filtering of the file list passed to epydoc.
Now only relax modules ending in *.py are processed.  That means that
all base directory scripts, including 'sconstruct', are excluded from
the API documentation.
    * API manual: More reliable parsing of the epydoc output to detect
non-wxPython issues.
    * FSF Copyright Validation configuration: Improvements to the
repository configuration section.  The different configurations can
now be chosen via a variable, rather than requiring code to be
uncommented.
    * FSF Copyright Validation script: Sorted the years for the
committer information output.  This makes it easier to read the file
and will help with compression.
    * FSF Copyright Validation script: A new file with all committer
information up to 2018.  This is to allow for much faster execution of
the FSFCV script, by only looking at the git log from the start of
2019.
    * FSF Copyright Validation script: Support for skipping the first
commit.  This is for truncated history, where for example the git
repository start date is set to a later date than the git repository
migration or initial SVN commit, when the committer information up to
a given date is read from a file.
    * FSF Copyright Validation script: Fix for tracking renames when
saved committer information is used.
    * Travis CI config: Execution of the FSF copyright validation
script as part of the testing.  This is set to run only on a new
Python 3.7 build job, simply to avoid unnecessary repetition.  All of
the git history needs to be fetched for the script to work, and the
script requires the pytz Python module.
    * FSF Copyright Validation script: Addition of a repository
configuration printout.  This is to help in debugging, as it is
otherwise not clear where the source of the copyright information
comes from.
    * Release checklist document: Rewrote the 'preparation'
instructions for Travis CI.  All previous manual checking is now
performed automatically by Travis CI for each push to the GitHub
mirror repository.


Bugfixes:
    * relax GUI: wxPython-Phoenix 4.x fix to allow relax to start
again.  In the later wxPython versions, relax would not be able to
start either the GUI or any of the test suite due to a new error
"wx._core.PyNoAppError: The wx.App object must be created first!".
This was not present in wxPython-Phoenix 3.  The Relax_icons class (a
wx.IconBundle derived class) is no longer instantiated on import.
    * Structure loading: Fix for bug #12, the acceptance of the
invalid '#' character in molecule names.  See
https://sourceforge.net/p/nmr-relax/tickets/12/.  A simple check has
not been added to the load_pdb() and load_xyz() functions of the
internal structural object in lib.structure.internal.object.  This
ensures that the '#' character can never be set as the molecule name,
independently if it was taken from a file name or set via the
set_mol_name arguments of the structure.read_pdb or structure.read_xyz
user functions.
    * Bruker DC: Complete redesign of the backend to support reading
newer (or older) file versions.  This fixes bug #13
(https://sourceforge.net/p/nmr-relax/tickets/13/), the failure of
reading newer Bruker DC files.  The backend has been resigned so that
the relax library produces a complex Python object representation of
the Bruker DC results file.  This object now stores all of the data
present within the Bruker DC file.  The design is more flexible as
precise column ordering no longer matters.
    * Fix for bug #14, the freezing of user functions in the GUI.  See
https://sourceforge.net/p/nmr-relax/tickets/14/.  The user functions
freeze if an error occurs that is not a RelaxError, with the mouse
pointer stuck on the busy cursor.  These non-RelaxErrors are now
caught and manually dealt with by the GUI interpreter.  Like all GUI
freezing bugs, this was introduced with the huge GUI speed up prior in
relax 4.1.0.  These also only to appear to be a freeze, but it is
actually the failure to update and show the relax controller combined
with not turning off the busy mouse cursor.
    * GUI bug fix: Avoidance of the numpy depreciation of '== None'.
This deprecation causes the GUI to fail with recent numpy versions.
    * Relaxation dispersion: Protection of all of the MODEL_PARAMS_*
variables from modification.  These are now only used with
copy.deepcopy().  This removes a number of bugs in which the lists,
which should be read-only, are permanently modified by the addition of
'r1'.  The system tests add 'r1' and then the unit tests subsequently
fail.  This would also be an issue if an experiment without the 'r1'
parameter is analysed after one with that parameter, without
restarting relax.
    * Relaxation dispersion bug fix: The 'r1' parameter was missing
from the nested parameter algorithm.  This is the nesting_param()
function of the specific_analyses.relax_disp.model module.  The 'r1'
parameter must be treated differently from the other model parameters,
just as the 'r2*' parameters are.
    * Dispersion auto-analysis: Bug fix for the plotting of the R1
parameter.  The plotting relied on the insertion of the 'r1' parameter
into the read only MODEL_PARAMS_* variables of
lib.dispersion.variables.  Now the Model_class class from
specific_analyses.relax_disp.models is being used to dynamically
determine the parameters of the model.