Opened 9 years ago
Closed 9 years ago
#5832 closed defect (worksforme)
CPLSetErrorHandler causes segmentation fault still open?
| Reported by: | sergiosegala | Owned by: | warmerdam |
|---|---|---|---|
| Priority: | normal | Milestone: | |
| Component: | default | Version: | unspecified |
| Severity: | normal | Keywords: | |
| Cc: |
Description
Hi.
We are using a Centos + Apache httpd + PHP + mapserver stack and are seeing dying httpd childs that seem related to gdal.
On one machine (call it "v") we have
Centos (6) + Apache httpd (2.2.15) + PHP 5.4.32 (from ius) stack + mapserver (5.6.9) and we're using
gdal-libs-1.9.2-3.el6.x86_64 gdal-python-1.9.2-3.el6.x86_64 gdal-1.9.2-3.el6.x86_64
compiled from source.
The children die with messages:
child pid xxxx exit signal Segmentation fault (11), possibile coredump in /xxx
and no other error message are present in the httpd error log.
when we have a look at the cores with gdb we obtain the attached stacktraces (repeating lines cut). There seems to be a bad recursion there that leads to a stack overflow, I guess.
On another machine (call it "c") we have a very similar setup (mapserver is 6.4.1 here), with
gdal-libs-1.11.1-2.el6.x86_64 gdal-1.11.1-2.el6.x86_64 gdal-python-1.11.1-2.el6.x86_64
also compiled from source.
Here the children die with messages:
child process xxxx still did not exit, sending a SIGTERM
and we see the following log line:
FATAL: CPLGetTLSList(): pthread_setspecific() failed!
In this case the stack traces cannot be obtained from the core dump because the stack is corrupt (we obtain a message that memory cannot be accessed at some location).
While these two crashes appear to be different here, we've found this bug report:
(appearantly closed) that very suspiciously seems a combination of our two: it has the loop in the stack trace and the reference to pthread_setspecific().
We suspect this issue is still there even though it appears to have been closed for gdal 1.9...
Can you help us?
-- Sergio Segala
Attachments (1)
Change History (3)
by , 9 years ago
comment:1 by , 9 years ago
comment:2 by , 9 years ago
| Resolution: | → worksforme |
|---|---|
| Status: | new → closed |
No feedback, no obvious way how to reproduce. Closing
Hum, not obvious what might go wrong. You would at least need to recompile GDAL 1.11 --enable-debug to remove optimizations and get debug symbols so that the stack trace might be useful.