Re: IDL Segmentation Fault [message #50517 is a reply to message #50398] |
Wed, 04 October 2006 16:34   |
Karl Schultz
Messages: 341 Registered: October 1999
|
Senior Member |
|
|
On Wed, 04 Oct 2006 15:09:13 -0700, mankoff wrote:
> Karl Schultz wrote:
>> On Wed, 04 Oct 2006 09:06:26 -0700, mankoff wrote:
>>
>>> David Fanning wrote:
>>>> Bill Anderson writes:
>>>>
>>>> > Has anyone run into this problem? I have running IDL 6.2 student
>>>> > edition for the past month with no problems on a MacPro Book (MacOS X
>>>> > 10.4) notebook. This weekend I updated my MacOS to 10.4.8 and now when
>>>> > I try to run idl from X11, it crashes with a "Segmentation Fault".
>>>> >
>>>> > /Users/wmandersonjr/idl
>>>> > IDL Student Edition 6.2, Mac OS X (darwin ppc m32). (c) 2005, Research
>>>> > Systems, Inc.
>>>> > Segmentation fault
>>>> > /Users/wmandersonjr/
>>>>
>>>> You might find this article helpful:
>>>>
>>>> http://www.dfanning.com/tips/maccrash.html
>>>>
>>>> Also, we discussed this problem right here less than two weeks
>>>> ago, it seems to me. Check the archives.
>>>>
>>>
>>>
>>> I've read that article and searched the archives. I'm using IDL 5.6 on
>>> a Intel MacBook. X11 programs work fine, but IDL is segfaulting
>>> nonetheless.
>>>
>>> If DISPLAY is :0.0 then xeyes works and IDL segfaults
>>>
>>> If DISPLAY is 0:0 then IDL will run but provides error messages when
>>> trying to access the display, like so:
>>>
>>> _X11TransSocketINETConnect() can't get address for 0:6000: No address
>>> associated with nodename
>>> % Unable to connect to X Windows display: 0:0
>>>
>>>
>>> Any advice?
>>>
>>> This just started yesterday when I upgraded to 10.4.8.
>>>
>>> -k.
>>
>> Yes, I've been looking at this recently, and have collected some hopefully
>> useful information.
>>
>> IDL PPC binaries of any release now appear to crash when running on an
>> Intel Mac under Rosetta emulation with OS X 10.4.8. The problem didn't
>> occur with OS X 10.4.7. The crash logs indicate a crash in libXt. Since
>> the *only* change made to the environment where I observed all of this was
>> upgrading from 10.4.7 to 10.4.8, I am inclined to suspect the Apple update
>> to 10.4.8.
>>
>> Some web searching reveals that there are more than a few problems with
>> the 10.4.8 upgrade on Intel Macs. One person reported a similar crash
>> with Matlab 7.1 under the same conditions. The report indicated that the
>> Matlab crashes started after the upgrade to 10.4.8 was applied. Same
>> situation, same problem, I think. One could also extend this thought to
>> the possibility that any PPC binary X11 application running under Rosetta
>> on an Intel based OS X 10.4.8 might crash.
>>
>> Therefore IDL users of versions prior to 6.3 probably should not upgrade
>> their Intel Macs to 10.4.8.
>>
>> IDL 6.3 users should really get the Intel binaries for IDL that are
>> available from ITTVIS. Not only would this relieve this particular 10.4.8
>> problem, but the Intel binaries run faster on the Intel Mac.
>>
>> The PPC-based hardware does not seem to be adversely affected by this OS X
>> upgrade.
>>
>> I'm going to go ahead and report this problem to Apple, and all we can do
>> is hope that they'll take the time from the Leopard schedule to release a
>> patch for this seemingly defective OS X upgrade.
>>
>> That's the best I can do for now,
>>
>> Karl
>
> Thanks for taking this up with Apple. Any luck just swapping out libXt
> with the previous version? I don't have access to my backup disk for
> another 10 days while I travel so I cannot test this.
I wouldn't even try it, because even if it worked, nobody would want
to do anything important in that sort of environment. I would also think
that more than just that one module is involved. And I didn't see any
crashes when using a VNC server on the same box, so the X server itself
may be part of the problem.
And by the way, I did copy the /usr/X11R6/bin/xclock binary from a ppc OS
X machine and it crashed on the intel 10.4.8 machine, as I suspected it
would. That's the way I filed the bug with Apple.
Karl
|
|
|