chicken-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Chicken-users] heap size or macro problem


From: Rick Taube
Subject: Re: [Chicken-users] heap size or macro problem
Date: Mon, 5 Nov 2007 06:30:04 -0600

The first thing to check is whether you perform any Scheme->C calls
with callbacks into Scheme: if these are not properly declared as "safe",
since this always (sooner or later) leads to trouble.

wow thanks so much you have saved us days of work with this tip! this is exactly what we are doing. a user scheme function gets added to a JUCE scheduling/scheme thread, which then does a callback on the fuctino at the correct time. during that callback the user's (scheme) function will certainly call into C again, for example sending midi notes to the output port. originally we were calling the user's function (the 'closure' var) from the scheme thread using C_callback like this:

{
      double nexttime ;
      C_word *elapsed_ptr;
      C_word elapsed_word;
      C_word closure;
      closure = CHICKEN_gc_root_ref(gcroot);
      elapsed_ptr = C_alloc(1);
      elapsed_word = C_flonum( &elapsed_ptr, time - start);
      C_save( elapsed_word );
      nexttime = C_c_double( C_callback(closure, 1));
}

i *think* this code was working but the downside is that every user funcino has to be wrapped with error handling code to avoid crashes under the callback... so we replaced that first version with one that uses a define-external

(define-external (run_process (scheme-object closure) (double elapsed)) double
  ( closure elapsed))


that get called like this:

nexttime = run_process( CHICKEN_gc_root_ref(closureGCRoot), time - start);

which -- if it worked -- i think would be a better solution becuse the we can put the error handling code in the external point instead of inside every user function, eg:

(define-external ...)
  (with-no-errors ( closure  elapsed))


if these are not properly declared as "safe",
since this always (sooner or later) leads to trouble.

so clearly we are not doing this right because nothing is declared as safe however I dont see in the docs how to declare the run_process as safe, can you help me out? or should we just use the first version and wrap each clouser with error proctection boilerplate?

thanks!
--rick




reply via email to

[Prev in Thread] Current Thread [Next in Thread]