Re: griddata segmentation fault

[Chenjie Gu]

Thanks very much, Thomas and Kai.

I found out that if I use gcc-4.1.2 (with -O2), it works fine. (I was using gcc-4.3.2)

Also, after I recompiled both qhull and octave using gcc-4.3.2, actually, griddata works, but the surf command is not working properly. And even if I just type "figure", it shows an error about drawnow saying cannot call gnuplot...

Maybe there's some other problem on my side.

Chenjie

On Tue, Dec 30, 2008 at 2:03 AM, Kai Habel <address@hidden> wrote:

Thomas Weber schrieb:

> On Tue, Dec 30, 2008 at 08:44:48AM +0100, Kai Habel wrote:
>
>> Chenjie Gu schrieb:
>>
>>> Thanks, David.
>>>
>>> If x,y,z are column vectors, it works on one of my intel machines.
>>> (which encountered an error without segmentation fault before.)
>>>
>>> However, still I got a segmentation fault on one of my amd64 machine
>>> (running 32-bit linux).
>>>
>>> Is there anything to do with the processor?
>>>
>>> Chenjie
>>>
>> Chenje,
>>
>> griddata depends on delaunay triangulation. Does delaunay work at all on
>> this machine(s) ?
>>
>> If not, I suspect you have been hit by this bug[1]. If that is the case
>> recompile qhull without optimization
>>
>
> It seems Fedora has a patch:
> https://bugzilla.redhat.com/show_bug.cgi?id=432309
>
>       Thomas
>

Thomas,
thanks for the pointer. I did a quick recompile with the srpm [1] and
the patch above [2] and can confirm that qhull works again with the
standard compiler flags (-O2).

Kai

[1] http://www.qhull.org/download/qhull-2003.1-1mdk.src.rpm
[2] https://bugzilla.redhat.com/attachment.cgi?id=296416