[whatwg] Possible bug in the way the spec about worker GC behavior

[Ian Hickson]

On Wed, 9 Oct 2013, Ehsan Akhgari wrote:
>
> Right now the spec says[1] "Whenever a Document object is discarded, it 
> must be removed from the list of the worker's Documents of each worker 
> whose list contains that Document.".  If I'm reading this correctly, 
> this implies that the worker object should be alive by the time that the 
> document gets discarded, which is not what Gecko implements.

Generally the spec doesn't try to imply anything, so if you are reading it 
as implying something, you're probably reading it wrong. :-)

Once a worker has shut down (its closing flag is set to true, its event 
loop is destroyed, and the "run a worker" algorithm has terminated), 
changing which documents are in the list of worker's Documents does 
nothing. So if you cleaned up the worker because it stopped being 
protected, then it doesn't matter what happens to the list of the worker's 
Documents. But, to make this clearer, I've adjusted the spec to explicit 
empty the list at the end of "run a worker".


On Thu, 10 Oct 2013, Ehsan Akhgari wrote:
> 
> It's even worse than that, we GC the worker object if we can prove that 
> it will not have any outstanding work to do in the future.

That's what the spec allows, right. (Exactly what counts as outstanding 
work is precisely defined, see the definition of "protected worker". That 
definition might need some work, though.)


> Well, removing a document from the worker's list of documents to me 
> implies that the worker object is not GCed, which implies that UAs 
> cannot GC worker objects until the document is discarded.

You definitely don't want to read that far between the lines.

Removing a document from the worker's list of documents means nothing more 
than that. The worker can have been GC'ed, in which case, nothing 
interesting happens.


On Thu, 10 Oct 2013, Andrew Wilson wrote:
> >
> > It's even worse than that, we GC the worker object if we can prove 
> > that it will not have any outstanding work to do in the future.
> 
> I suspect that would break in the case of SharedWorkers.

SharedWorkers are explicitly exempted from this.


I made a couple of changes to the spec here, hopefully they help (one 
change makes the text clearer to avoid the issue in this thread, the other 
change fixes the lifetime requirements to allow suspended workers to get 
GC'ed, since it doesn't make sense to require those to live...):

   http://html5.org/tools/web-apps-tracker?from=8328&to=8330

(Sorry for marking them editorial instead of normative.)

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'