Re: warnin about 'info' is a bug in eigs-base.cc

[Daniel J Sebald]

On 08/30/2012 04:05 PM, John W. Eaton wrote:
> On 30-Aug-2012, John W. Eaton wrote:
>
> | On 30-Aug-2012, Jordi Gutiérrez Hermoso wrote:
> |
> | | I guessssss that I shooooouuld instead refactor and untangle the
> | | sparse code. Do you insist I do that? It seems like a stupid
> | | refactoring effort, but I guess we'll have to do it sooner or later.
> |
> | I was initially very confused about what was going on with the double
> | use of info.  At the very least, I think that tricky code like that
> | deserves a comment to explain it.
> |
> | I'm not insisting that you completely refactor and fix up that code.
> | But I would rather have a "force" argument for the SparseCHOL
> | constructor, and to not overload info for this purpose.  I'd also
> | prefer to avoid having using "nargout" in liboctave functions, as that
> | is not really a concept that applies there.  If you don't want to do
> | it, I will.
>
> I checked in the following change:
>
>    http://hg.savannah.gnu.org/hgweb/octave/rev/94cdf82d4a0c
>
> This change doesn't fix the cut and paste problems

That's fine; not too much with one changeset.


>  but I think it does
> clarify the code somewhat by avoiding the confusing overloading of the
> meaning of info.  It should also fix the uninitialized warning.

It compiles here without warning.  "force" is a little vague, but so is 
the code.

I don't know how to test the sparse Cholesky routines, off hand, but the 
changes all look good--references, bool options, etc.

Dan