KPEDBG_HDL_POP_FCPTR

  • From: Vít Špinka <vit.spinka@xxxxxxxxxxxx>
  • To: ORACLE-L <oracle-l@xxxxxxxxxxxxx>
  • Date: Mon, 17 Dec 2012 23:53:17 +0100

Hi list,
have anyone seen something like this?

We have a Perl+C application connecting to Oracle; it uses DBD::Oracle 
and OCI (although it does not seem to actually use OCI in this case). It 
runs more or less fine on Linux, Windows etc. And if it crashes, it has 
signal handlers installed to create a stack trace, collect various 
diagnostics etc. So far so good.

However, on  Solaris on Sparc, we on some machines get instead just 
short error messsage like "KPEDBG_HDL_POP_FCPTR" and then the program 
dies, without invoking any of the signal handlers. And truss shows it 
actually got some signal that caused it to do this.
My guess is that this is kpeDbg() function that can be sometimes seen on 
other platforms when the crash is in Oracle libraries - it prints a 
neatly formatted call stack.

So the question is - have you ever seen something like this? Do you know 
of any way how to get at least some information as to nature of the 
issue? E.g. let the signal get to the Perl+C program or get the call stack.

Version info: Instant Client 11.2.0.3, database: 10.2.0.5, 11.2.?
Tried: adding bequeath_detach=yes and DIAG_SIGHANDLER_ENABLED=FALSE to 
sqlnet.ora, but to no avail.

TIA
Vit


--
//www.freelists.org/webpage/oracle-l


Other related posts: