Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
Sender: trawick@rdu88-250-035.nc.rr.com
To: jim@jaguNET.com
Cc: dev@apr.apache.org, apr-cvs@apache.org
Subject: Re: cvs commit: apr/locks/unix crossproc.c locks.c proc_mutex.c
References: <200204042120.QAA06138@devsys.jaguNET.com>
From: Jeff Trawick <trawick@attglobal.net>
Date: 04 Apr 2002 18:42:43 -0500
In-Reply-To: <200204042120.QAA06138@devsys.jaguNET.com>
Message-ID: <m3bsczm3nw.fsf@rdu88-250-035.nc.rr.com>
Lines: 41
User-Agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/20.7
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

Jim Jagielski <jim@jaguNET.com> writes:

> Jeff Trawick wrote:
> > 
> > "sem_open("/ApR.whatever", O_CREAT, 0644,1)" returns ENOSYS on Linux
> > (2.2.12 kernel).  I'll try to add an autoconf check for that.
> 
> Yeah... right now almost all the APR checks are just for existance
> and not implementation... we should really fix that for those
> that we expect to work (all the shared mem and semaphore and thread
> stuff), but that's a lot of work (but the right way to do it).

I wouldn't worry about this.  It was silly for Linux to have a stub
that returns ENOSYS.  Handling this for everything we check for would
just be code bloat.

> > Beyond the issue mentioned above, why is a new lock mechanism
> > suddenly so high in the priority list?  Aren't we throwing away a lot
> > of past experience (1.3 and 2.0)?
> > 
> 
> Posix sems are new. The logic being that they are "better" than SysV
> sems for those platforms that support both.

I don't see any logic there.  Some platforms will just add posix sems
to check off an entry in a table.  Adding later doesn't mean it is
implemented better.  It may in fact mean that it is implemented in
terms of something that already existed (i.e., slight overhead for new
feature, which incidentally no important programs use since it has
never existed so nobody bitches). 

Also, we don't know how these things work in practice.  What if (wild
hypothesis) Solaris has it but the default kernel config allows only a
handful of them?  Then we start getting PRs from a bunch of the people
who did --disable-threads on Solaris (and therefore aren't using
pthread mutexes) and don't understand that "Out of space on device"
means their system isn't configured for enough posix sems?

-- 
Jeff Trawick | trawick@attglobal.net
Born in Roswell... married an alien...