Hey,
The lock tries to be re-entrant when called from the same thread in some
cases.
The following is ok:
read lock
read lock
read unlock
read unlock
write lock
read lock
read unlock
write unlock
The following is not ok:
read lock
write lock
write unlock
read unlock
The following probably will lock it up too:
write lock
write lock
write unlock
write unlock
Keeping these semantics is vital, since a lot of code depends on them.
On Fri, 2004-02-06 at 00:23, Brian Hawkins wrote:
> Boy did I open a can of worms. My ReadWriteLock class was lacking in two areas. It does not allow for nested read locks or a read lock inside of a write lock.
>
> So I created a test suite to hammer my read write lock. I added the above two features and then ran it along side the read write lock from jedit. This is where I found a bug in the jedit version. It allowed a read lock to occur when a write lock was already in place.
>
> Being very proud of what I had done I put my class in the jedit code and compiled. It worked great until I hit the undo button in jedit and the whole thing locked up. Looks like that bug is a feature ;)
>
> It seems to me that allowing for nested locks is not a good design and tends to complicate the code. What does everyone else think?
>
> Also are there any unit tests written for the jEdit source? Would anyone object if I started adding them to the build?
>
> Any way for anyone who cares to see what I've come up with I've attached my ReadWriteLock class
>
> Brian
--
Slava Pestov
|
