Re: [griddata]:__delaunay__ function

[Martin Helm]

Am 15.11.2011 10:38, schrieb Petr Gajdos:
> On Mon, Nov 14, 2011 at 11:23:35PM +0100, Martin Helm wrote:
>> Am 14.11.2011 22:58, schrieb Thierry Laroche:
>>> I also try this on opensuse 11.4, and I obtain the same result.
>>> I will first fill a bug report and if it's too long I should compile
>>> as you suggest.
>>> keep in touch
>>>
>>> Thank you
>> When you go for a compile from its original source look also at this
>> page it will help you
>>
>> http://www.schoeps.org/home/2010/01/install-octave-on-opensuse/
>> it is for version 11.3.
>>
>> Make sure that you have qhull5 (!) installed before you build
> Hi,
>
> thanks Thierry to pointing me to this thread, I am on this list, but
> not reading everything; feel free to ping me more loudly with future 
> problems with openSUSE's octave.
>
> If I understand correctly I should downgrade qhull to less recent
> version? I have added qhull into science repository not long ago as 
> octave dependency, so it should be done easily. Without patch we 
> have against octave qhull6 gets not detected at all.
>
> Thanks for any help,
> Petr
>
The problem is really that your patch makes octave compile with the
newer qhull 2011 but the resulting binaries give wrong results. So there
seems to be a deeper incompatibility on the binary level which cannot be
solved by changing the include files.
With qhull 2010 the results are correct.
I found a discussion on this list from 30.08.2011 where builds for mac
and mingw also have problems with qhull 2011 by giving wrong results
after compiling octave against it, so it seems to be in general that
octave is incompatible with qhull6 whatever trick one applies to make it
compile.
I can confirm the same for me when I compile my octave 3.4.3 directly
from the downloaded sources and also the current development version on
openSUSE 11.4 against qhull5 it works, but against qhull6 (patching the
seek for the right include files) it gives just wrong results.