[Ecls-list] sig11 in cl_boot()

[Juan Jose Garcia-Ripoll]

2007/1/16, Stanislaw Halik <sthalik+ecl at tehran.lain.pl>:
> I disabled all other pthread-using libraries, made sure it isn't linked
> in anymore, linked against libecl, yet it still crashes.
> libecl is now the first library being linked:
> LIBS =  -L/tmp/ecl/lib/ecl -lecl -lncurses  -lssl -lcrypto -lnsl
> -lresolv -lcrypt -lm
> [...]
> Looks like my libecl wasn't built with pthread support after all.
> cl_boot() is the first thing run in main, no threads are created before
> its execution.

As I said before, it seems to me that the garbage collector is missing
something. Otherwise there is no explanation why intern() is being
called with a package value of 0x0 (i.e. null pointer).

I can only offer you two things. The first one are two a quick test
whether this has to do with the detection of static roots. First, in
the debugger, make sure that stacks_scanner() is called at least once.
This is the routine responsible for marking the lisp variables as
used.

If positive, remove the line GC_no_dls=1 in alloc_2.d That should
force the garbage collector to look for static variables in all
shared libraries. You should also try removing the lines
        old_GC_push_other_roots = GC_push_other_roots;
        GC_push_other_roots = stacks_scanner;
        GC_start_call_back = (void (*)())finalize_queued;
and see what happens.

The final thing is that, if this is really a Debian/x86/32bit box, I
could debug the object files for you. Make a tarball of your program
with debugging information and sources and send it to me by private
email. It should not be very difficult to track down the problem
since, as you say, ECL is the first block of code to be started and
the crash happens shortly after initializing the core.

Regards,

Juanjo

-- 
Dpto. de Fisica Teorica I, Fac. de CC Fisicas, Universidad Complutense,
Ciudad Universitaria s/n Madrid 2804 (Spain)
http://teorica.fis.ucm.es/~jjgarcia/