On Mon, Nov 10, 2014 at 1:59 PM, Seth Fowler <s...@mozilla.com> wrote:
> In bug 1091921, we got support for using Maybe with the Auto helpers for
> Mutex and Monitor - things like MutexAutoLock and MonitorAutoEnter. This
> supports a pattern for optionally acquiring a RAII resource that I first saw
> used in the JavaScript engine, and which I’ve found very useful since. For
> anyone who hasn’t seen it, the basic pattern looks like this:
>
>> Maybe<ExpensiveRAIIResource> resource;
>> if (resourceIsNeeded) {
>> resource.emplace();
>> }
>
> This constructs an ExpensiveRAIIResource on the stack only if
> |resourceIsNeeded| is true.
>
> So bug 1091921 lets us use this pattern with Mutexes and Monitors. I’m sure
> people will find lots of situations where this is useful, and indeed it’s
> already being used in some places.
>
> Any time you have parallelism, though, you need to exercise caution. I
> encourage anyone who wants to use this to start by adding assertions to their
> code like Mutex’s |AssertCurrentThreadOwns| or Monitor’s
> |AssertCurrentThreadIn| anywhere they have methods that expect another method
> to do their synchronization for them. That’s good practice in any case, and
> will help ensure that you don’t make a mistake when using Maybe in this way.
>
> Enjoy!
>
> - Seth
> _______________________________________________
> dev-platform mailing list
> dev-platform@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-platform
Since we're on the subject, I'll point out that
AssertCurrentThreadOwns is debug only (looking at you b2g, with your
almost complete lack of debug testing) and AssertNotCurrentThreadOwns
is unimplemented and for documentation purposes only.
- Kyle
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
