Handling of old registry format

[githubkusi]

PRQ #94 introduced the record field toolboxSubfolder. A previous registry record (probably saved in the prefs) doesn't have that field. Provide the user with a proper error message and a solution

The code still breaks if the user has an old record in his prefs (which happened with a colleague of mine), but at least it is more clear now. We could try to handle this correctly (being backward-compatible), but I don't think it's worth the risk & time. Or what do you think?

The best solution for future changes would be to introduce a version field in the record generated by tbToolboxRecord().

[githubkusi]

please merge first PR #103, otherwise the wrong commits are assigned to THIS pull request

[DavidBrainard]

I'm a little confused here. Many toolbox records do not have the toolboxSubfolder field, and yet they do not seem to be breaking. What is the exact situation that causes the failure?

[DavidBrainard]

Nope, this PR doesn't fix the TbGitAndSanityTest suite - four still broken on my machine, but as noted elsewhere these were not broken for me prior to PR 102. Will try to dig in a little.

[DavidBrainard]

The (or at least, one) issue occurs in the call to parser.parse in tbParsePrefs, as called from tbAddToolbox -> tbWriteConfig -> tbParsePrefs. The value of persistantPrefs is empty, and this makes the parser unhappy when called with arg (varagin{:}) because there is no key associated with empty variable persistantPrefs.

I have not delved into how the new persistantPrefs is implemented, but maybe the problem will be immediately obvious to you?

[githubkusi]

that's strange. So you say getpref('ToolboxToolbox') is empty for you? Even then, an empty persistent pref struct does not get fed into parser.parse. So I don't understand, I'm sorry. Could you please post a callstack?

[DavidBrainard]

runtests TbGitAndSanityTest/testAddOneAtATime:

================================================================================ Error occurred in TbGitAndSanityTest/testAddOneAtATime and it did not run to completion.

Error ID:
---------
'MATLAB:InputParser:ParamMustBeChar'
--------------
Error Details:
--------------
Error using tbParsePrefs (line 74)
Expected a string scalar or character vector for the parameter name.

Error in tbWriteConfig (line 18)
prefs = tbParsePrefs(tbGetPersistentPrefs, varargin{:});

Error in tbAddToolbox (line 51)
tbWriteConfig(config, persistentPrefs, prefs);

Error in TbGitAndSanityTest/testAddOneAtATime (line 77)
                results = tbAddToolbox(tbGetPersistentPrefs, ...

[DavidBrainard]

And indeed, getpref('ToolboxToolbox') is empty. I run the TbTb in its default configuration, so I don't think any preferences ever get set.

[DavidBrainard]

I don't think my diagnosis above is quite right. What is true is that it errors as in the call stack above, and that persistentPrefs is empty.

[githubkusi]

Indeed, if setpref('ToolboxToolbox') is empty, it also fails here: You see the same failures when running all tests on TbGitAndSanityTest?

Failure Summary:

 Name                                      Failed  Incomplete  Reason(s)
=========================================================================
 TbGitAndSanityTest/testAddOneAtATime        X         X       Errored.
-------------------------------------------------------------------------
 TbGitAndSanityTest/testSubfolderPath        X         X       Errored.
-------------------------------------------------------------------------
 TbGitAndSanityTest/testTwoSubfolderPaths    X         X       Errored.
-------------------------------------------------------------------------
 TbGitAndSanityTest/testHook                 X         X       Errored.

[DavidBrainard]

Yes, I'm pretty sure it was those four (I am at another machine now, but there were definitely four failures from TbGitAndSanityTest, and only four failures.)

[githubkusi]

should be fixed with #105