Subj : Re: question: memory allocated by JS_Enumerate().
To   : jp
From : Brendan Eich
Date : Fri May 06 2005 12:36 am

jp wrote:
> Did a search in the list, didn't find anything.
> 
> I have been hunting a memory leak for the past 3 days. Now i suspect 
> that the memory allocated by JS_Enumerate() will not be release unless 
> you call JS_DestroyIdArray(). So here's my questions:
> 
> 1. Do I need to use JS_DestroyIdArray() to release the pointer returned 
> by JS_Enumerate()?


Yes, of course -- but the stupid API doc doesn't say so.  Still, it's 
pretty clear from the signatures.


> 2. It seems, if I don't use JS_DestroyIdArray() to release the pointer, 
> the memory will not be released even after I destroy the context and 
> runtime. Is that true?


Yes.


> 3. If 2 is true, I think it is very easy for users to forget to call 
> JS_DestroyIdArray(). At least the document should emphasis it.


You're right, I'll fix it.


> 4. Is there any other situation similar to this where I should use extra 
> caution? I was really scared by this. :) It took my team 3 days to 
> figure it out.


The manually allocated and freed, or created and destroyed, data 
structures in the API are JSRuntime, JSContext, JSIdArray, 
JSExceptionState, and -- unless you wrap a script with an object using 
JS_NewScriptObject, JSScript.


> 5. It would be nice to have a way to find out how much memory the 
> runtime is using.


It would, but it's not as easy as it seems.  Instrumenting a wrapped 
malloc may be the only way, and even then some strings handed off to the 
GC heap may not be counted properly.

/be

.