scm_with_continuation_barrier returning #f on throws ?

scm_with_continuation_barrier returning #f on throws ?

[Cedric Cellier]

Hello !

With guile-1.8.7, scm_with_continuation_barrier does not seam to return
SCM_BOOL_F on error. I did this :

static void *load_file(void *filename)
{
    scm_c_primitive_load((char const *)filename);
    scm_force_output(scm_current_output_port());
    return NULL;
}

...

SCM res = scm_with_guile(load_file, "a_file_that_doesnt_exist");

Thus, scm_c_primitive_load displays an ERROR message and throws out.
But scm_with_guile, ie. the underlying scm_with_continuation_barrier,
returns 0x0.

Is the bug in the doc, in the code or once again in my head ?

Re: scm_with_continuation_barrier returning #f on throws ?

[Andy Wingo]

Hi Cedric,

On Mon 06 Sep 2010 13:20, Cedric Cellier <rixed@happyleptic.org> writes:

> SCM res = scm_with_guile(load_file, "a_file_that_doesnt_exist");

This statement is ill-formed, because you shouldn't have references to
SCM objects outside of Guile. This is indicated by scm_with_guile's
return value, which is void* and not SCM.

The docs about scm_with_continuation_barrier are correct; it does indeed
return #f on a throw. But scm_with_guile uses
scm_c_with_continuation_barrier, which is different, because it returns
a C value, a void*. The error case there has it return NULL.

Regards,

Andy
-- 
http://wingolog.org/

Re: scm_with_continuation_barrier returning #f on throws ?

[rixed]

Sorry, my bad, I just can't read code (hopefully I'm paid to /write/ some :-))