Re: Writable lock nested in read-only lock

Hi Brian,

Brian Smith wrote:
> Isn't this a problem no matter what, even if I start the outer 
> transaction as a read-write transaction?

No without introducing an external lock. Then of course you can have 
deadlock even without using MDR and transactions at all :).
Usually when you introduce an external lock and use it by two different 
threads the threads know of each other.
This is not the case with threads using MDR. There can be different 
independent modules working on the same instance of MDR. Now when one 
thread of one module decides to read and the thread of another module 
decides to read, they both are granted the read access. Now one of the 
modules decides to start writing inside of the read. And the other 
module decides to do that too - now you come to a deadlock, because the 
first module is waiting for the second module to leave the read 
transaction and vice versa. It is not feasible to allow that, because it 
is practically impossible to prevent such a deadlock.
It could be solved by releasing the read lock when waiting for write 
lock, but that would lead to unexpected results.

> For my application, having READ_ONLY_PROMOTABLE would be perfect but 
> READ_WRITE_DEMOTABLE (which should never cause any deadlock problems, 
> right?) would be "good enough".

OK, I will think about adding the DEMOTABLE transactions when I will 
revisit the transactions API.
Martin