Comment 120 for bug 604635

Comment on attachment 596637
Part 2, Implement wakelock interfaces

One crucial thing which is missing here, leaving aside the IPC discussion above, is how we automatically cancel wake locks.

I think all wake locks a document acquires should be void when the document is unloaded (as in, onunload). Some wake locks should probably be canceled earlier than that -- for example, the screen wake lock should probably be canceled if the document which requested the lock becomes invisible (e.g., I switch away from that tab). We can deal with these special cases later, but we should deal with the general case here.

A lot of this code is going to change with the changes to part 1, so I'm not doing a full review here. But skimming this patch, one thing to note is that we don't follow the C convention of declaring all our variables at the start of a block. Instead, we declare immediately before use.

For example, this

+ PRUint32 lock;
+ bool isHeld;
+
+ isHeld = mLockHashtable.Get(aTopic, &lock);

becomes

+ PRUint32 lock;
+ bool isHeld = mLockHashtable.Get(aTopic, &lock);