Subj : Re: JS_AddRoot & finalize problem
To   : netscape.public.mozilla.jseng
From : anton muhin
Date : Fri Jul 23 2004 02:47 pm

Hello, Brendan!

Brendan Eich wrote:
> Perhaps something was added to the array later?
No. Proto is constructed and I immediatley leave jsshell session.

> GC_MARK_DEBUG and js_liveThingToFind would tell the answer, if you're 
> willing to re-compile with the former defined, and set the latter at 
> runtime to point to the object (|obj| in your code shown last time) that 
> never gets collected.
I beg your pardon for small correction in case someone else would read 
your postings: not |js_liveThingToFind|, but |js_LiveThingToFind| (at 
least in my version of the sources).

I tried it and finished with:

003294a8 object 00000000 Array via root(Array)

Here come questions:

003294a8---an address of created Array.

00000000---sources name it rootThing. Do I interpret it correctly as 
private data of the object that leaks, in this case an instance of Array?

Array---obvious, class name

root(Array)---I suppose it's a kind of a root dedicated to Array? Maybe 
the one I introduced with JS_AddRoot while constructing.

If I trace my prototype object I got:

<quote note = "a little bit reformated">
00329490 object 0042665C MFInt32 via 
root(Array).__proto__(Array).__proto__(Object).__parent__(global).MFInt32(Function).prototype(MFInt32).

00329490 object 0042665C MFInt32 via 
root(Array).__proto__(Array).__proto__(Object).__parent__(global).MFInt32(Function).prototype(MFInt32).
toString(Function).__parent__(MFInt32).
</quote>

Couple of comments: toString is the only method I define for my MFInt32 
class.

I'd be wrong but it looks like that the prototype doesn't get collected 
as it has a pointer to rooted array as its *private* data. If it's so, 
I'm a little bit astonished, for it seems counter-intuitive to me. Is 
private-data slot treated differently by GC?

with the best regards,
anton.

.