I have had really bad luck with signals. They can be “lost” in more than one 

I thought this was the far more reliable and correct way… at least from that 
point of view.




From: Nathan Nobbe [mailto:quickshif...@gmail.com] 
Sent: Monday, May 02, 2011 5:01 PM
To: Jeremy Greene
Cc: php-general@lists.php.net
Subject: Re: [PHP] semaphore release before acquire warning


On Mon, May 2, 2011 at 2:57 PM, Jeremy Greene <jer...@zeevee.com> wrote:

        I am getting a warning when calling sem_release() before (the same php
        script) calls sem_acquire(). I am doing this because it's a "signal" to
        another process. The other process (which happens to be C program) has
        done, or will do, a semop() to acquire the semaphore/signal. The actual
        data transfer is through shared memory.
        It does all functionally work quite nicely, but given that I'm getting
        the warning and that there doesn't seem to be any discussion of this at
        least in this list's archive maybe I'm putting a square peg into a round
        hole... or at least there's a rounder peg available.
        I did look into disabling the warning, but that got me more concerned
        since it seemed like a frowned upon thing to do and even more of a
        performance hit.
        The irony is that I'm using shared memory (and signals) exactly for
        performance reasons L


perhaps try pcntl_signal() to signal the c program rather than sem_release().



Reply via email to