[Ecls-list] [sagemath] On early_signal_error, cl_boot and threads

[Julien Puydt]

Hi,

The current sagemath compiles ECL with disabled threads, which is a 
problem to package it in debian.

The problem is that when compiled with threads, things start to 
crash(well, assert), because early_signal_error gets called.

What happens is that sage and the gc are using signals and for some 
reason, one is raised before the call to cl_boot.

In fact, sage's initialization of ECL looks like:

ecl_set_option(ECL_OPT_SET_GMP_MEMORY_FUNCTIONS,0);
ecl_set_option(ECL_OPT_SIGNAL_HANDLING_THREAD, 0)
cl_boot(1,arv)


I'm at loss what to do to fix the issue: how does ECL have 
early_signal_signal_error set as signal callback even before cl_boot was 
run!?

Thanks,

Snark on #sagemath