mTotalBytes had the meaning of "the sum of all elements in the vector",
but was reset to 0 at the start of a transaction. The meaning is now
changed to "total number of bytes in the vector that belong to the
current transaction".
An LLFolderView is added as child to LLScrollableContainerView,
but also adds the LLScrollableContainerView to it's mScrollContainer.
As a result, when scrolling inside a LLFolderView the event is passed to
the mScrollContainer, which then passes it first on to it's children
(see the "Bad UI design" remark in the code), causing an infinite loop.
This patch breaks that loop for those objects that have a
mScrollContainer: LLFolderView and LLContainerView.
This adds a counter and a 'dead' flag to the data stored in the linked
list. The counter counts the number of iterators (still) pointing to an
element and the dead flag is set to indicate the element was erased when
iterators are still pointing to it. As a result, iterators are NEVER
invalidated. Of course, such elements are subsequentially skipped when
iterating over the list. Assertions protect against dereferencing an
erased iterator, but incrementing or decremention still works: it is
still well-defined what the next (non erased) element is, assuming the
element wasn't erased (yet), but would be erased delayed - or assuming
the iterator would have been incremented (decremented) in advance to
erasing the element.
This is in fact much safer, because the only way to invalidate an
interator in the first place (in this code) is by calling removeChild,
which *already* has an assert never to remove a child that is being
iterated over (by means of the mInDraw flag).
This class can be used as RWLOCK parameter to AIThreadSafe to check that
data is only accessed by a single thread (like
AIThreadSafeSingleThreaded) AND is never write locked when a read or
write lock was already obtained (by the same thread). It doesn't
actually lock anything, it just keeps track if the "lock" was obtained
before. The use case is to check if STL containers aren't being used
(read or write locked) by a calling function when additional write
access is necessary, as write access might invalidate iterator that
point to the container (where the previous lock was taken).
This allows passing a different type than the default AIRWLock for
debugging purposes.
The signature of the class used for RWLOCK should be:
struct RWLock {
// Default constructor.
RWLock(void);
// Block until it's safe to read the data.
// high_priority is a hint that granting this thread the read lock is more important than granting another thread a write lock.
void rdlock(bool high_priority = false);
// Release the obtained read lock.
void rdunlock();
// Block until it's safe to write to the data.
void wrlock();
// Release the obtained write lock.
void wrunlock();
// Convert the obtained write lock into a read lock.
void wr2rdlock();
// Block until it is possible to convert the obtained read lock into a write lock.
void rd2wrlock();
// Return true when a read or write is obtained and subsequent calls to release the lock are expected.
bool isLocked() const;
};
