GCC C++ Issue: pure virtual method called

[Stefan Arentz]

I have a threaded C++ application that produces a 'pure virtual method called'
error about 3 out of 10 times while starting up the application. The error is
confusing because this should be a compile-time error instead of runtime.

My question is how I can trap an error like this so that I can look at a stack
trace from the debugger. I tried several things like setting a breakpoint on
__cxa_pure_virtual but none of these hooks seem to get called. It is difficult
to find the correct bit of documentation about these GCC 'internals'.

This is my GCC version:

Reading specs from /usr/libexec/gcc/darwin/ppc/3.3/specs
Thread model: posix
gcc version 3.3 20030304 (Apple Computer, Inc. build 1492)

I know that this is beta software but I also have the same problem with older
versions of GCC like 2.95 and 3.2.

S.

[Mike]

>

> I have a threaded C++ application that produces a 'pure virtual method called'
> error about 3 out of 10 times while starting up the application. The error is
> confusing because this should be a compile-time error instead of runtime.
>
> My question is how I can trap an error like this so that I can look at a stack
> trace from the debugger. I tried several things like setting a breakpoint on
> __cxa_pure_virtual but none of these hooks seem to get called. It is difficult
> to find the correct bit of documentation about these GCC 'internals'.
>
> This is my GCC version:
>
> Reading specs from /usr/libexec/gcc/darwin/ppc/3.3/specs
> Thread model: posix
> gcc version 3.3 20030304 (Apple Computer, Inc. build 1492)
>
> I know that this is beta software but I also have the same problem with older
> versions of GCC like 2.95 and 3.2.

My guess is that it's a pointer error (a programming error). Some pointer points
at the wrong place in the vtab and causes a call to a method which cannot be
called.

[Stefan Arentz]

"Mike" <nospam@nospam.com> writes:

> >
> > I have a threaded C++ application that produces a 'pure virtual method called'
> > error about 3 out of 10 times while starting up the application. The error is
> > confusing because this should be a compile-time error instead of runtime.
> >
> > My question is how I can trap an error like this so that I can look at a stack
> > trace from the debugger. I tried several things like setting a breakpoint on
> > __cxa_pure_virtual but none of these hooks seem to get called. It is difficult
> > to find the correct bit of documentation about these GCC 'internals'.
> >
> > This is my GCC version:
> >
> > Reading specs from /usr/libexec/gcc/darwin/ppc/3.3/specs
> > Thread model: posix
> > gcc version 3.3 20030304 (Apple Computer, Inc. build 1492)
> >
> > I know that this is beta software but I also have the same problem with older
> > versions of GCC like 2.95 and 3.2.

>
> My guess is that it's a pointer error (a programming error). Some pointer points
> at the wrong place in the vtab and causes a call to a method which cannot be
> called.

Yeah, probably combined with some weird timing issues. It could be many places, that
is why I need a stack trace :-)

S.

[Mike]

> > > I have a threaded C++ application that produces a 'pure virtual method called'

> > > error about 3 out of 10 times while starting up the application. The error is
> > > confusing because this should be a compile-time error instead of runtime.
> > >
> > > My question is how I can trap an error like this so that I can look at a stack
> > > trace from the debugger. I tried several things like setting a breakpoint on
> > > __cxa_pure_virtual but none of these hooks seem to get called. It is difficult
> > > to find the correct bit of documentation about these GCC 'internals'.
> > >
> > > This is my GCC version:
> > >
> > > Reading specs from /usr/libexec/gcc/darwin/ppc/3.3/specs
> > > Thread model: posix
> > > gcc version 3.3 20030304 (Apple Computer, Inc. build 1492)
> > >
> > > I know that this is beta software but I also have the same problem with older
> > > versions of GCC like 2.95 and 3.2.

> >
> > My guess is that it's a pointer error (a programming error). Some pointer points
> > at the wrong place in the vtab and causes a call to a method which cannot be
> > called.

>
> Yeah, probably combined with some weird timing issues. It could be many places, that
> is why I need a stack trace :-)

Have you looked in the crash log?

[Benjamin Riefenstahl]

Hi Stefan,

Stefan Arentz <stefan.arentz@soze.com> writes:

> My question is how I can trap an error like this so that I can look
> at a stack trace from the debugger. I tried several things like
> setting a breakpoint on __cxa_pure_virtual but none of these hooks
> seem to get called.

"br __cxa_pure_virtual" works fine here to trap this (GCC 3.3, Apple
build 1435). __cxa_pure_virtual() calls abort() here, so you can also
enable core dumps, run the program, reproduce the error and than run
gdb on the code dump.

You probably already know that this error is caused by calling pure
virtual methods from the constructor or destructor of the base class.
It must be an indirect call, because GCC will issue an error about
direct calls.

benny