Re: [Boost-users] [interprocess] named mutex clean up

Subject: Re: [Boost-users] [interprocess] named mutex clean up
From: peter_foelsche_at_[hidden]
Date: 2009-04-08 13:20:00

> My intention is just providing portability. "Create and unlink" named 
> resources are usual in UNIX systems and I find them quite useful. And 
> shared_memory is modeled after POSIX shared memory, that's why I wante
> d 
> to provide a more homogeneous behaviour with the library. This 
> portability was also requested by some users, but who knows, this migh
> t 
> be a step in the wrong direction.
> 
> Why do you think this unlink behaviour is worse than "file in use" err
> or 
> in windows?

I don't see any need to be able to remove mutexes which are in use.
The sense of a named mutex is that only one mutex with a particular name can exist on a system.

I still remember having to clean up named system objects (shared memory I think) by and then on SOLARIS.
I don't like the idea of being able to confuse running processes.

Somewhere in the SOLARIS documentation was a (quite ridiculous) message, which said (from memory):

"Obligatory file locks are dangerous. Should a process hold a obligatory file lock on a system file, it can halt the entire system."

Or:

"Software updates should be done when the system is not much in use."

Taking into account that a process usually crashes when the used executable or shared library file is being written to, I found this obvious acceptance of crashes not acceptable.