Problem with ARPACK?

[Alexander Mamonov]

A small update to the "clear all" crashing Octave issue. It appears
that this is a manifestation of behavior observed in 3.0.x series,
when certain octave-forge packages caused Octave to crash after "clear
all". I reinstalled Benjamin's 3.2.0 MinGW binary without any
packages, and "clear all" no longer crashes Octave. I will try to
identify the package that causes problems. Unloading packges with "pkg
unload all" doesn't help, the package must be uninstalled.
However, ARPACK-related crash

> eigs(sparse(1:10, 1:10, 1:10, 10, 10));
> clear all
> error: stdin is not a tty!

is still present with all octave-forge packages uninstalled.


On Thu, Jun 11, 2009 at 11:57 AM, Alexander Mamonov <mamonov at gmail.com> wrote:
> So what is the best way to add tests that detect the ".oct function
> call + clear all" bugs?
> Is something like
>
> %! test
> %! svds(sparse(1:10, 1:10, 1:10, 10, 10));
> %! clear all;
>
> allowed in tests? Is it OK for a failed test to crash instead of
> generating an error?
>
> BTW, in Benjamin's build of 3.2.0 that is now available on sourceforge
> I noticed a bug, that was supposed to be fixed by
> http://hg.savannah.gnu.org/hgweb/octave/rev/1500d0197484
> i.e.
>
>>backend('gnuplot');
>>clear all
>
> crashes Octave. This is really strange, since that changeset fixed the
> problem for then current sources, and as I can see in the 3.2.0
> tarball, it is included in the 3.2.0 release. Can somebody confirm?
>
> Regards,
> Alex
>
> On Wed, Jun 10, 2009 at 1:37 PM, John W. Eaton <jwe at octave.org> wrote:
>> On 10-Jun-2009, Alexander Mamonov wrote:
>>
>> | Well, first the bug must be confirmed by someone else. Maybe it's just
>> | a problem with my particular build. If the problem is there, a new
>> | test should definitely be added.
>> | I have once attended a talk by Cleve Moler (the guy who created
>> | Matlab), and he mentioned, that every bug that is fixed in Matlab
>> | becomes a test in their test suite (the full suite takes days to run).
>> | It would be nice if Octave community adopted this practice.
>>
>> So, go back through the bug reports and submit patches that add
>> tests.  No one is preventing you from doing this.  Our community still
>> seems somewhat short of active contributors given that Octave is a
>> relatively large project.  So I'm sure that it would be appreciated if
>> you (or anyone else) contributed to Octave by adding tests.
>>
>> jwe
>>
>