[CLN-list] cln 1.1.9

[Richard B. Kreckel]

On Tue, 15 Mar 2005, Richard B. Kreckel wrote:
> On Tue, 15 Mar 2005, Andrej Ricnik-Bay wrote:
>
> > > First, say 'run'.  Let it crash.  Then, say 'bt'.
> > Ouch :) ... I feel thoroughly stupid now ;0
> > (gdb) bt
> > #0  0x40051b6e in cl_inc_pointer_refcount (pointer=0x0) at object.h:205
> > #1  0x40057917 in cln::cl_gcobject::inc_pointer_refcount (this=0x4017d36c)
> >     at object.h:317
> > #2  0x400578ca in cln::cl_gcobject::_as_cl_private_thing (this=0x4017d36c)
> >     at object.h:501
> > #3  0x40057833 in as_cl_private_thing (x=@0x4017d36c) at object.h:506
> > #4  0x400589a3 in cl_FF (this=0x4017dad0, x=@0x4017d36c) at ffloat_class.h:72
> > #5  0x400cae5b in cln::cl_I_to_FF (x=@0xbffff3ac)
> >     at float/ffloat/elem/cl_FF_from_I.cc:34
> > #6  0x400cb0fb in cln::cl_RA_to_FF (x=@0xbffff3ac)
> >     at float/ffloat/elem/cl_FF_from_RA.cc:36
> > #7  0x4012f397 in __static_initialization_and_destruction_0 (__initialize_p=1,
> >     __priority=65535) at real/format-output/cl_fmt_scaleexp.cc:46
> > #8  0x4012fa28 in global constructors keyed to
> > _ZN3cln21format_scale_exponentERKNS_4cl_FE () at
> > real/format-output/cl_fmt_scaleexp.cc:331
> > #9  0x4013acb5 in __do_global_ctors_aux ()
> >    from /home/andrej/source/cln_1.1.9/src/.libs/libcln.so.3
> > #10 0x40042629 in _init ()
> >    from /home/andrej/source/cln_1.1.9/src/.libs/libcln.so.3
> > #11 0x4000aaae in call_init () from /lib/ld-linux.so.2
> > #12 0x4000a906 in _dl_init_internal () from /lib/ld-linux.so.2
> > #13 0x400007d5 in _dl_start_user () from /lib/ld-linux.so.2
>
> Hmm?  I wonder whether cl_fmt_scaleexp were initialized before the
> cl_FF_globals?  Just to make double sure, please send me the files
> cln/include/intparam.h and cln/include/floatparam.h.  Also, you compiled
> with -g -O2, right?  Please check whether those are indeed the compiler
> flags when you compiled the library!

Andrej: I've been discussing exactly the same problem with Ralf Wildenhues
in private and it turned out that it comes from turning optimization off.
Indeed, on March 9 you wrote:

> Sorry, yes, it's x86, compiled with P4 optimizations
> (env has CXXFLAGS and CFLAGS=-march=pentium4 -msse2)

There's no -O flag.  I completely missed that error back then.  We've now
added a warning to the configuration script for such cases.  I assume this
issue is finally resolved now.

Regards
  -richy.
-- 
Richard B. Kreckel
<http://www.ginac.de/~kreckel/>