I put some inline comments in. I'll redo the samples after we hammer
this out.
> -----Original Message-----
> From: Warren Young [mailto:mysqlpp@stripped]
> Sent: Thursday, October 25, 2007 10:24 AM
> To: MySQL++ Mailing List
> Subject: Re: Sample files -- BadQuery w/Errnum patch part 4
>
> Jim Wallace wrote:
> > + deadlock1 and deadlock2: Tests detecting a deadlock between
> > + two transactions, using the
> BadQuery::what_errnum() value
>
> I've looked at these, and have some concerns:
>
> 1. The database creation stuff belongs in resetdb.
Ok
>
> 2. Why does the first example need the loop? The second pass
> isn't actually needed to make it deadlock, is it?
Not to get the deadlock, but to have it succeed, it must do the second
pass since the first pass throws the exception, rolls back the
transaction. In the exception handling, the what_errnum() is checked,
which will not be the same as errnum() as the output shows.
>
> 3. If the new tables are created and populated ahead of time
> by resetdb, it seems to me that the two programs are so close
> to each other that it should be possible to have just one
> program, which you run twice. I don't quite understand the
> nature of the deadlock, but maybe if you just add second
> cin.getline() call between the two SELECT queries, so the
> first is sitting and waiting halfway through when you start
> the second?
I thought about doing it and actually had #ifdef's originally. The
problem is that deadlock2 hangs waiting for the lock 1. Once deadlock1
tries to get lock 2 (which deadlock2 has), it immediately gets the
deadlock exception, and rolls back, releasing lock1 so deadlock2
completes ok. And after deadlock2 completes, deadlock1 can retry and
succeed.
I couldn't see how to do it since deadlock2 hangs, and deadlock one
needs to wait for it to get in the hung state to trigger the deadlock.
Threads would have worked, but I didn't want to get into x-platform
issues.
If you have some suggestions, I'd be happy to try them.
