Reliability and shutdown

[ael]

On Sun, May 13, 2012 at 11:15:15AM -0700, Tracy Reed wrote:
> I'm surprised these problems have persisted for so long. The offlineimap
> website and docs actually make it look like someone really cares about this
> project. But the persistent bugs suggest otherwise or that hardly anyone uses
> this piece of software.

It is an extremely useful piece of software and it is very actively
maintained.

But concurrency issues are subtle and without formal methods it is
exceedingly hard to diagnose such problems. As in this case, they tend
to show up in rare cases, and developers generally don't see them. If
they did, they would find a solution - often something that works
for the vast majority.

So almost by definition, only a few people see them in rare
circumstances. Most people using the software are very happy.

That the problems have a root in race conditions can only be
speculation. There is also the issue of all the different
implementations of the imap protocol and whether they conform
to an imprecise specification.

I run a number of different machines with widely varying speeds and 
architectures. offlineimap behaves differently on all and the apparent
deadlocks may only occur on 1 or 2 of the systems. But which often
changes with the offlineimap release. This, and the fact that others see
such problems, all smacks strongly of race problems. But I could be
wrong and have no right to criticize (which I am not doing). I hope
one day to have time to investigate, but I do not expect it to be 
easy :-(

ael