[webkit-dev] [webkit-changes] [56710] trunk/WebCore

[Eric Seidel]

Agreed.  BCC'd webkit-changes moving to webkit-dev.

http://trac.webkit.org/changeset/38411
mentions the GCC bug I was referring to.  I've never seen the Radar so
I don't know what it was about exactly.

It sounds like DEFINE_STATIC_LOCAL is still needed for other reasons tough.

We should consider adding the iPhone threading issue to the list of reasons at:
http://trac.webkit.org/browser/trunk/JavaScriptCore/wtf/StdLibExtras.h#L31

Thanks for the heads-up.

-eric

On Mon, Mar 29, 2010 at 12:07 PM, Darin Adler <darin at apple.com> wrote:
> On Mar 29, 2010, at 11:07 AM, Eric Seidel wrote:
>
>> Can we kill DEFINE_STATIC_LOCAL yet? It was originally to work around a GCC bug. However maybe there are reasons why we still need it these days?
>
> This seems like the wrong mailing list for an in-depth discussion of it.
>
> Nothing has changed that I am aware of. I’m not sure what GCC bug you are referring to.
>
>> Is it to make shut-down faster by preventing destructors from running?
>
> Lack of control of what thread the destructors run on was one major issue that caused us to do this. The threading issue created serious problems on the iPhone platform as atomic strings were destroyed on the wrong thread.
>
> The ability to get extra speed on shutdown by opting in to this mechanism is also a benefit.
>
> For all the ports that Apple uses, shutting down the library without shutting down the entire process is a non-goal.
>
>    -- Darin
>
>