On 2010-05-25, at 2:14 PM, Emit Sorrels wrote:

Hello,

I'm experimenting with Gambit as an embedded scripting engine (to replace a Lua-based one) inside a large application written in C. For the most part, the scm<->c interop is working well for my use-case.

I would now like to add a remote REPL for debugging and changing application logic on the fly. I added examples//web-repl.scm into my app and exposed start-repl-server within a c-define.

Surprisingly it works, but I notice that if I do ,q or ,qt the application segfaults. gdb bt is no help #0 0x00007ffff742c2b8 in siglongjmp () from /lib/libc.so.6 #1 0x00000000004c0f3c in ___call () #2 0x00000000004bff3e in start_repltest () #3 0x00000000004c0060 in main ()

How exactly do you start the Scheme code? Ideally you should model your main program along the lines of lib/_setup.c . The important part is to delimit the C code which calls Scheme code inside a ___BEGIN_TRY and ___END_TRY pair so that exceptions (such as the one which is used to terminate the process when you enter ,q) will be caught, i.e.:

___BEGIN_TRY scheme_proc(1,2,3); // call Scheme procedure (defined using c-define) ___END_TRY

Moreover, if you want to catch Scheme exceptions in all REPL threads created, then you should set the right flags in setup_params.debug_settings . So the main function of your program will be:

int main (int argc, char **argv) { ___setup_params_struct setup_params; ___setup_params_reset (&setup_params);

setup_params.version = ___VERSION; setup_params.linker = SCHEME_LIBRARY_LINKER; setup_params.debug_settings = (___DEBUG_SETTINGS_UNCAUGHT_ALL << ___DEBUG_SETTINGS_UNCAUGHT_SHIFT) | (___DEBUG_SETTINGS_ERROR_REPL << ___DEBUG_SETTINGS_ERROR_SHIFT);

___setup (&setup_params);

{ ___SCMOBJ ___err; ___processor_state ___ps = ___PSTATE;

___BEGIN_TRY

start_repltest();

___END_TRY }

___cleanup ();

return 0; }

Also, the application no longer responds to ctrl-c.

It turns out that there were a few bugs in the implementation of the telnet protocol in example/web-repl.scm . I've attached a corrected repltest.scm file which fixes the problem. Now each REPL created will have its own thread-group and a CTRL-C at a REPL will interrupt all threads in that thread group.

Let me know if that works out for you.

Marc