Re: 'make check' errors for Mac OS X

[Rik]

On 11/02/2013 03:50 PM, c. wrote:
> On 2 Nov 2013, at 23:13, Rik <address@hidden> wrote:
>
>> After resolving the conv errors, I think we need to take a look at the
>> other platforms--Mac and Windows--and see what failures we have there.  In
>> some cases it may be approriate to add ismac(), ispc() in the %!tests to
>> special case certain tests.
> On the mac I get the following:
>
> Summary:
>
>   PASS     12671
>   FAIL        13
>   XFAIL        3
>   SKIPPED      7
>
> out of the 13 fails, 7 are in conv, and 1 is bug 38236.
> The remaining 5 are as follows:
>
> 1 failure in chol-cc.tst which is a tolerance issue
> 1 failure in testing error messages in figure.m
This actually exposed an oversight.  I recoded figure.m to require that the
input argument be a scalar
(http://hg.savannah.gnu.org/hgweb/octave/rev/cfe5297e0f3b).  It should pass
now.
> 2 failures in datestr, due to uppercase/lowercase differences.
This turns out to be Apple's interpretation of the %p argument to
strftime.  Unlike every other UNIX-like OS, %p produces lower case am/pm
instead of uppercase ("Think Different", and make life harder for others
:().  I re-coded the tests so that they will pass on Mac or Linux now
(http://hg.savannah.gnu.org/hgweb/octave/rev/2ff47214de72).
> 1 failure in svds which is a tolerance issue
I'd like to see if other Mac users experience the same issue.  We could
increase the tolerance if it consistently fails by the same amount on
multiple different Mac systems.
>
> None of these looks very bad, I think the main issue on OSX is that syntax 
> highlighting crashes the GUI.
There seems to be some progress on the syntax highlighting issue.  I can
see that you and Torsten have been working on the issue quite a bit.  I'm
willing to continue to wait a few days in the hopes that this can be sorted
out.

--Rik