Subj : JS_ResumeRequest puzzle
To   : netscape.public.mozilla.jseng
From : Bob Kline <bkline@rksystems.com>
Date : Wed Mar 17 2004 11:46 am

I'm trying to apply the generous advice we've received over the past few 
days on making the Sabltron XSL/T package's use of SpiderMonkey 
thread-safe.  The latest snag is a compilation failure:

.....jsext.cpp(465) : error C2660: 'JS_ResumeRequest' : function does not 
take 1 arguments

Lines 464-466 have:

#ifdef JS_THREADSAFE
   JS_ResumeRequest(cx);
#endif

where cx is declared as JSContext*.  Here's what the documentation says 
for JS_ResumeRequest:

=====================================================================

JS_ResumeRequest 	Function
Summary
Restarts a previously suspended thread.
Syntax

void JS_ResumeRequest(JSContext *cx);

Description
When your application restart a previously suspended thread, 
JS_BeginRequest safely increments the thread counter for the JS engine 
runtime associated with a given context, cx. In order to increment the 
counter, this function first checks that garbage collection is not in 
process. If it is, JS_ResumeRequest waits until garbage collection is 
complete before locking the JS engine runtime and incrementing the 
thread counter. After incrementing the counter, JS_ResumeRequest unlocks 
the runtime if it previously locked it.

Notes
JS_ResumeRequest is only available if you compile the JS engine with 
JS_THREADSAFE defined. In a default engine compilation, JS_THREADSAFE is 
undefined.

=====================================================================

Can I get around this problem by avoiding JS_SuspendRequest and 
JS_ResumeRequest altogether, replacing:

     JS_BeginRequest(cx);
     JS_XXX();
     JS_YYY();
     JS_SuspendRequest(cx);
     BlockingOperation();
     JS_ResumeRequest(cx);
     JS_ZZZ();
     JS_EndRequest(cx);

with

     JS_BeginRequest(cx);
     JS_XXX();
     JS_YYY();
     JS_EndRequest(cx);
     BlockingOperation();
     JS_BeginRequest(cx);
     JS_ZZZ();
     JS_EndRequest(cx);

on the assumption that these two blocks of code are functionally 
equivalent to each other?  Or is my understanding of the request model 
still not incorporating all of the subtleties?

Also, one more further clarification.  Brendan corrected one of the 
things I said the other day, pointing out that it is not *necessary* to 
put the JS_GC() and JS_MaybeGC() calls outside of the request blocks. 
Does that also mean that it is not *allowed* as well, or is it safe to 
put the GC calls anywhere?

Thanks again!

Bob

.