Re: segfault when embedding the interpreter

[John W. Eaton]

On 10/12/2013 03:47 AM, c. wrote:
> Hi,
>
> With a recent version of the development sources
> whenever I try to run standalone programs embedding the interpretr, e.g. 
> examples/embedded.cc, I get a segfault.
>
> $ hg id
> 1e8f8900a041 tip @
> $ mkoctfile -g --link-stand-alone embedded.cc -o embedded
> $ ./embedded
>
>
> Below is the result of running the program in the interpreter
> and the backtrace of the segfault.
>
> GCD of [10, 15] is 5
>
> Program received signal EXC_BAD_ACCESS, Could not access memory.
> Reason: 13 at address: 0x0000000000000000
> 0x00000001024e4b1e in std::_Rb_tree_increment ()
> (gdb) bt
> #0  0x00000001024e4b1e in std::_Rb_tree_increment ()
> #1  0x000000010036e7c0 in symbol_table::erase_subfunctions_in_scope ()
> (gdb)
>
>
> last time this was working was a couple of weeks ago.
> I submitted this as a bug: https://savannah.gnu.org/bugs/index.php?40248
>
> Can anyone confirm this behaviour?

I see a different problem:

  coredump:2103> 
LD_LIBRARY_PATH=/usr/local/octave/dev/lib/octave/3.7.7+ ./embedded
  GCD of [10, 15] is 5
  *** Error in `./embedded': corrupted double-linked list: 
0x00000000017e23d0 ***

This failure doesn't happen for me if I exit the program with
"clean_up_and_exit (0)" instead of just returning.  But I don't think
that should be needed.

First, are you sure you are linking with the correct libraries?  Is it
possible that when you run the program it is being linked with some
older set of libraries?  If I do that, I get a segfault.

Either way, there does seem to be a problem because I don't think it
should be necessary to call clean_up_and_exit unless you need to
process atexit functions.  The other memory cleanup things that
clean_up_and_exit performs should be optional.

It might help if someone could bisect and find out when this problem
started happening.

jwe