[Top][All Lists]

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

Re: Octave-2.9.9 luinc failure

From: David Bateman
Subject: Re: Octave-2.9.9 luinc failure
Date: Tue, 03 Oct 2006 22:32:37 +0200
User-agent: Thunderbird (X11/20060921)

John W. Eaton wrote:
> On  3-Oct-2006, David Bateman wrote:
> | This is one of those bugs when you don't have all of the dependent
> | libraries installed. A possible fix though I'm unsure if it is complete
> | is the attached..
> OK, I already figured out the patch for luinc.cc.  I also checked in
> your change for MatrixType.cc, but with the check changed to p != 0
> instead of perm != 0, since it seems that on entry to this constructor
> perm will always be 0.
> With this change, I can build without sparse libraries and run the
> all the tests though some fail due to missing features.  This is not
> surprising to me, but I wonder if we should fix the tests so that they
> don't try to test things that can't possibly work due to missing
> functionality.  I'm not asking you to make the changes, just whether
> you think it is a reasonable thing to do.
> Thanks,
> jwe

Its certainly its reasonable, I did something like this in regexp.cc in
a number of places

%!test  # Capture replacement
%! if (!isempty(findstr(octave_config_info ("DEFS"),"HAVE_PCRE")))
%!   data = "Bob Smith\nDavid Hollerith\nSam Jenkins";
%!   result = "Smith, Bob\nHollerith, David\nJenkins, Sam";
%!   t = regexprep(data,'(?m)^(\w+)\s+(\w+)$','$2, $1');
%!   assert(t,result)
%! end

The number of places to do this however seems large... Also do we just
assume that if one sparse package in UFSparse is missing they all are or
should each test be disabled only if the specific functionality it needs
is missing? I.E. Do we support older versions of UFSparse in the tests.
If so the job becomes larger yet again to correctly disable the right


reply via email to

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