Re: 'make check' errors

[PhilipNienhuis]

Rik-4 wrote
> 11/2/13
> 
> So, far we continue to make all of our deadlines for the 3.8 release.  All
> of the language translation files have been updated and I committed the
> final patches today.
> 
> The next step in the release is to examine the errors resulting from 'make
> check' and determine what to do about them.  On a base linux system there
> is one failure resulting from bug #38236.  This won't get fixed for this
> release.  The other 7 errors are with the convolution function.  The first
> of these is a tolerance issue, but in the others the dimensions of the
> output and expected values differ.  This is so fundamental that I'm not
> sure the tests aren't wrong rather than the code.  Could someone with
> Matlab run the following?
> 
> --- Code ---
> a = rand (10, 10, 10);
> b = rand (3, 3);
> c = convn (a, b, 'full');
> size (c)
> d = convn (a, b, 'same');
> size (d)
> e = convn (a, b, 'valid');
> size (e)
> 
> a = rand (10, 15, 7, 8, 10);
> b = rand (4, 3, 2, 3);
> f = convn (a, b, 'same');
> size (f)
> g = convn (a, b, 'valid');
> size (g)
> --- End Code ---
> 
> 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.

A typical sample of what I get on Windows (MinGW) is posted in this thread:
http://octave.1599824.n4.nabble.com/Problems-with-the-Windows-build-tc4658593.html

Compared to "what we are used to on Windows" (~10 fails give or take a
couple) a few new FAILs came up; I suspect tolerance issues and one
(assert.m) is a simple case of different output format. Later this coming
week I'll try to look in the others.

Philip




--
View this message in context: 
http://octave.1599824.n4.nabble.com/make-check-errors-tp4658749p4658754.html
Sent from the Octave - Maintainers mailing list archive at Nabble.com.