Subj : Re: SEM_UNDO substitute on windows
To   : comp.programming.threads
From : Joe Seigh
Date : Mon May 23 2005 12:39 pm

On 23 May 2005 06:39:52 -0700, <deepesh.g@gmail.com> wrote:

> Thanks for your response.
> I am using semaphores to check if an application is already running
> (application tries to acquire lock on a semaphore and if times out, it
> must have been acquired by another instance of the application). In
> case if this application exits (May be user exits which is very good
> and may be because windows decides that it must be crashed now and
> thats where the requirement for SEM_UNDO) the semaphore operations it
> did must be undone so that another instance of application would be
> able to acquire lock on this semaphore and go ahead with execution. I
> would be happy to get any robust solution for this problem, but somehow
> I imagined that using semaphores would be a good solution (until I
> realize that SEM_UNDO is not supported on windows). Any input is
> welcome.

The fact that you're using synchronization means that the state of the
data matters.  What do you think the state of the application's data
is in if it crashes?

-- 
Joe Seigh

When you get lemons, you make lemonade.
When you get hardware, you make software.

.