octave-maintainers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: test failures in eigs and svds


From: Rik
Subject: Re: test failures in eigs and svds
Date: Mon, 09 Jan 2012 12:28:50 -0800

On 01/09/2012 10:34 AM, address@hidden wrote:
> Message: 2
> Date: Mon, 9 Jan 2012 18:01:41 +0000
> From: Carn? Draug <address@hidden>
> To: "John W. Eaton" <address@hidden>
> Cc: octave maintainers mailing list <address@hidden>
> Subject: Re: Octave 3.5.92 snapshot available for ftp
> Message-ID:
> <address@hidden>
> Content-Type: text/plain; charset="iso-8859-1"
>
> On 6 January 2012 21:02, John W. Eaton <address@hidden> wrote:
>> > Unless there are some serious problems with this snapshot, I would
>> > like to release it (or something very close to it) as 3.6.0 in the
>> > next few days. ?Even if you have been building Octave regularly from
>> > the Mercurial archive, it would help if you could build from the
>> > snapshot and report any problems.
> Hi
>
> While the build ran successfully in Debian Squeeze (current stable
> version), I have a few test failures when running "make check".
>
> Summary:
>
> PASS 9973
> FAIL 29
>
> There were 2 expected failures (see fntests.log for details).
>
> The failures are in the functions eigs (all 27 tests fail), svds
> (fails 2 of 3 tests). I have attached the fntests.log file. It's all
> because the eigs is not available (guess I should build the ARPACK
> library that is no longer being distributed with octave).
>
> I'll try to get the library installed, build it again, and report any
> problems (or hopefully success). But anyway, there should be some test
> to skip this tests and some warning after running configure if the
> library is not found.


The issue here is simple, but the solution might require a bit of coding.  The problem is that some of the failing tests require both the ARPACK library and the UMFPACK library or CHOLMOD library.  Unfortunately the %!testif sequence only accepts a single argument so only one library can be checked.  The preferred way to handle this would be to extend the testif processing in test.m to handle multiple arguments.  As a start, however, these failing tests should probably all be changed over to "%!testif HAVE_ARPACK" since that is the most important library for the svds and eigs functions.

--Rik


reply via email to

[Prev in Thread] Current Thread [Next in Thread]