reassign 860405 libocamlbricks-ocaml-dev
affects 860405 marionnet

On 16/04/2017 16:22, Chris Lamb wrote:
marionnet: segfaults after installation, not usable

If I rebuild the package I can still reproduce the segfault. A
quick attempt at getting a backtrace results in:

(gdb) bt full
#0  0x0000555555942a9c in camlCamlinternalFormat__make_printf_62299 () at
#1  0x00005555557eb741 in camlInitialization__entry () at
#2  0x00005555557169f9 in caml_program ()
#3  0x0000555555990e18 in caml_start_program ()
#4  0x00007ffff5ec4166 in gobject_init () at ././gobject/gtype.c:4450
        info =
          {class_size = 1, base_init = 0x0, base_finalize = 0x555555990fc0 
<__libc_csu_init>, class_init = 0x555555715cd0 <_start>, class_finalize = 
0x7fffffffeb00, class_data = 0x0, instance_size = 0, n_preallocs = 0, instance_init = 
0x55555597f548 <main+12>, value_table = 0x0}
        node = <optimized out>
        type = 140737488349104
        env_string = <optimized out>
#5  0x00007ffff5ec4166 in gobject_init_ctor () at ././gobject/gtype.c:4487
#6  0xba67aa8906c52ccb in  ()
#7  0x0000000000000000 in  ()

Thanks for this. It shows where the problem is: basically, ocamlbricks (which is a library used by marionnet) performs (via a function "f_") an Obj.magic on a format string, assuming it is internally a string, which is no longer true since ocaml 4.02.0 (4.02.3-3 has been uploaded to unstable on 2015-10-08). Marionnet has been broken since then.

I've checked in the latest trunk of ocamlbricks, and it is still broken:

And I've checked in the latest trunk of marionnet, it still calls f_:

So the problem still exists upstream. I will submit a bug there.



Reply via email to