Re: scm_init_guile when null threads

[Ken Raeburn]

On Apr 1, 2006, at 21:48, Oleg Parashchenko wrote:
> Neil Jerram <address@hidden> wrote:
> > Could you easily rework the program to use scm_boot_guile or
> > scm_with_guile instead?
>
> No, scm_boot_guile doesn't return, and I can't guess what  
> scm_with_guile
> does. But looking at sources, I found the function  
> scm_i_init_guile. This
> might be exactly what I need, I'll try later.

You call scm_with_guile, and give it a function to call that  
manipulates Guile objects or calls Guile functions; it causes the  
thread to enter "guile mode", calls the function, leaves "guile mode"  
and returns to the caller.  (You can also pass a pointer to whatever  
associated data you want, so you don't have to make your function  
rely on global variables.)

Since it controls entry and exit to your function, its ability to  
figure out where the "guile part" of the stack of that thread (needed  
for garbage collection) starts is much greater than with  
scm_init_guile.  If you use scm_init_guile, it needs to figure out  
where the thread's stack's base is (yes, this may be easier on some  
platforms when configured for null threads), because under that  
interface, you could pop back up a few stack frames before invoking  
some Guile routines and storing objects into stack slots.  Using  
scm_with_guile, your automatic variables with SCM objects won't get  
stored on the "wrong" side of the stack pointer that scm_with_guile  
has recorded.

Ken