[ZOO-Discuss] ZOO and segfaults

Sun May 16 17:49:42 PDT 2010

Hi All,

Thanks for your help Gerald and Luca, adding an [env] section with a PythonPath has fixed my problem. This is a little odd as I can see the PythonPath is set to "." by default in the debug trace when no PythonPath is set, but I am happy it's working now!

Should I be logging these segfault cases as bugs somewhere so they can be addressed at some stage in the future?

Next problem:

This link:

http://shilpa.media.osaka-cu.ac.jp/zoo/?request=Execute&service=WPS&version=1.0.0&Identifier=Distance&DataInputs=InputEntity1=Reference@xlink:href=http://carto.languedoc-roussillon.ecologie.gouv.fr/webservices/wfs/diren_general/%3FVERSION%3D1.1.0%26service%3DWFS%26request%3DGetFeature%26typename%3DZnieff1%26maxfeatures%3D1;InputEntity2=Reference@xlink:href=http://carto.languedoc-roussillon.ecologie.gouv.fr/webservices/wfs/diren_general/%3FVERSION%3D1.1.0%26service%3DWFS%26request%3DGetFeature%26typename%3DZnieff2%26maxfeatures%3D1

Does not work with the default Distance.zfg from Subversion on my server. When I debug this is because a segfault is caused (due to the VALUE being null I think?). Would it be possible to get the set of zfcg files from the shipla server? The Bounding box example with the xlink works, so I can reverse engineer each zfcg  if I need to.

Thanks for your help so far,

Cheers,
James Sewell
LISAsoft Developer
+61 (3) 8680 3200 / +61 414 688 892

From: Gérald Fenoy [mailto:gerald.fenoy at geolabs.fr]
Sent: Sunday, 16 May 2010 12:15 AM
To: James.Sewell at lisasoft.com
Subject: Re: [ZOO-Discuss] ZOO and segfaults

Hi James,
sorry for my late answer.

I'm not sure to know how to solve your issue, but can you tell me what is the PYTHONPATH set in your main configuration file in the specific env section if you get any ?

Indeed, it seems that there is some problem if you don't set the PYTHONPATH in the env section of your main configuration file unfortunately.

Please try to add the following section at the end of your main configuration file corresponding to your specific Python installation :

[env]
PYTHONPATH=<YOUR_PYTHONPATH>

Please let me know if this solve your issue.

Hope to hear from you,
best regards,

Le 14 mai 2010 à 09:20, James.Sewell at lisasoft.com<mailto:James.Sewell at lisasoft.com> a écrit :

Hey all,

After having worked out my previous problem (segfaults caused by an older version of CURL) I am faced with a new problem. I am now getting segfaults caused by  calls to free(). I get these problems when trying to run an Execute (I am testing with the python hello world at the moment).

A portion of the backtrace when debugging with gdb is below. It looks like calling free on python_path is the culprit.

*** glibc detected *** ./zoo_loader_capture: free(): invalid pointer: 0x000000000042515b ***
..
#4  0x00007ffff4c5be53 in free () from /lib/libc.so.6
No symbol table info available.
#5  0x00000000004117a3 in zoo_python_support (main_conf=<value optimised out>, request=0x631830, s=0x642150,
    real_inputs=<value optimised out>, real_outputs=0x7fffffff0820) at service_internal_python.c:178
        m = 0x632e30
        inputs = 0x651a60
        outputs = <value optimised out>
        tmp = <value optimised out>
        python_path = 0x42515b "."
        pModule = <value optimised out>
        pFunc = <value optimised out>
...

I have replicated this behavior on a CentOS 64bit machine and a Ubuntu 32bit machine.  The Java hello world runs fine, although it looks like some of the other processes (Distance etc..) have other problems causing segfaults for other reasons (a call to xmlNewText which seems to be in the libxml2 library in the case of Distance).

I can stop the free errors by setting MALLOC_CHECK_ =0, but that doesn't seem like a very good solution.

Can anyone shed any light on why this is happening? Has anyone managed a successful install of the subversion head which can process python? I can run all these requests from shilpa.media.osaka-cu.ac.jp/zoo/?, so I am also curious about what version this is running. I can see some differences on this box to the head (the Distance process only requires one Input as an example).

I am more than happy to provide more debugging output if anyone would like some.

Cheers,
James Sewell
LISAsoft Developer
+61 (3) 8680 3200 / +61 414 688 892

_______________________________________________
Zoo-discuss mailing list
Zoo-discuss at gisws.media.osaka-cu.ac.jp<mailto:Zoo-discuss at gisws.media.osaka-cu.ac.jp>
http://gisws.media.osaka-cu.ac.jp/mailman/listinfo/zoo-discuss

Djay
Just a ZOO Tribe member
gerald.fenoy at geolabs.fr<mailto:gerald.fenoy at geolabs.fr>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/zoo-discuss/attachments/20100517/cd7ee12e/attachment.html>