Re: [OctDev] arpack: eigs crash with octave 3.0.5 compiled with --enable-64

[Alexander Barth]

On Wed, Jun 24, 2009 at 8:18 AM, Jaroslav Hajek<address@hidden> wrote:
> On Tue, Jun 16, 2009 at 3:56 PM, Alexander
> Barth<address@hidden> wrote:
>> On Tue, Jun 16, 2009 at 3:21 PM, Jaroslav Hajek<address@hidden> wrote:
>>> On Tue, Jun 16, 2009 at 3:06 PM, Alexander
>>> Barth<address@hidden> wrote:
>>>> On Mon, Jun 8, 2009 at 4:32 PM, Alexander
>>>> Barth<address@hidden> wrote:
>>>>> On Sun, Jun 7, 2009 at 3:51 PM, David Bateman<address@hidden> wrote:
>>>>>> Jaroslav Hajek wrote:
>>>>>>>
>>>>>>> Do we? In fact, I think the Fortran standard requires the default
>>>>>>> REAL, INTEGER and LOGICAL to be of equal size (numeric storage unit).
>>>>>>> So, only some combinations of these flags are conforming in general. I
>>>>>>> do not think Octave needs to be bulletproof in this regard, at least
>>>>>>> not until a real need arises from practice. When you're altering the
>>>>>>> sizes of default types, you're often breaking standard-conformance,
>>>>>>> and you're therefore supposed to know what you're doing. Anyway, a
>>>>>>> common practical advice is to avoid LOGICALs in external interfaces at
>>>>>>> all and just use INTEGERs.
>>>>>>>
>>>>>>> cheers
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> Then the pgf90 compiler with the -i8 flag is not standards compliant. 
>>>>>> Given
>>>>>> that, then yes just using octave_idx_type is the right thing
>>>>>>
>>>>>> D.
>>>>>
>>>>>
>>>>> Sorry, forgot the attachement.
>>>>>
>>>>> Alex
>>>>>
>>>>
>>>> Hi all,
>>>> Will my patch by applied? Is it now OK?
>>>>
>>>> Thanks
>>>> Alex
>>>>
>>>
>>> Please include also a ChangeLog entry, I'll apply it then.
>>>
>>> regards
>>
>> As a ChangeLog entry, you may use:
>>
>> Use octave_idx_type for parameters of type LOGICAL in ARPACK
>>
>> Thanks and regards!
>> Alex
>>
>
> Ooops, sorry for the delay, I accidentally unstarred this. Could you
> please supply the ChangeLog entry in the proper format? See ChangeLog
> files for details:
> http://hg.savannah.gnu.org/hgweb/octave/file/8bec23396924/liboctave/ChangeLog
> I can do it eventually if you won't, but probably later ...
>

What about this? (maybe the first line is not necessary)

2009-06-24  Alexander Barth <address@hidden>

 * eigs-base.cc (
EigsRealSymmetricMatrix,
EigsRealSymmetricMatrixShift,
EigsRealSymmetricFunc,
EigsRealNonSymmetricMatrix,
EigsRealNonSymmetricMatrixShift,
EigsRealNonSymmetricFunc,
EigsComplexNonSymmetricMatrix,
EigsComplexNonSymmetricMatrixShift,
EigsComplexNonSymmetricFunc): Use octave_idx_type for parameters of
type LOGICAL in ARPACK

Thanks and regards,
Alex