Just wanted to confirm that this also occurs on Mac OS 10.6 using version PHP 5.3.0.

Sorry, but your problem does not imply a bug in PHP itself.  For a
list of more appropriate places to ask for help using PHP, please
visit http://www.php.net/support.php as this bug system is not the
appropriate forum for asking support questions.  Due to the volume
of reports we can not explain in detail here why your report is not
a bug.  The support channels will be able to provide an explanation
for you.

Thank you for your interest in PHP.

$ cat 51160.php 
<? 

pcntl_signal(SIGPIPE, function() { exit; });

passthru ("/bin/sh 51160.sh | head -n1");

I'm confused.  If the exec family of functions is designed to execute something as if executed from a shell, and the command executes properly from a shell, but not in php... how is this not a php bug?

I also have this issue. I've tried to test a simple example on several of our web hosting machines and all of them have this issue present.

Simple test. Create a php file (test.php) with this content:
==================
<?php
shell_exec('ls | ls | ls | ls' );
==================
and then run: php -f test.php 2>&1
you should get the error like: ls: write error: Broken pipe
almost every time you run that command.

I'm not sure what's the reason for this, but I'm pretty sure it's the bug, since the bash/sh doesn't ever raise that error no matter how many ls'es we put into the pipe.

Just to say, my problem somehow got misteriosly solved. I don't know exactly what was the culprit, but I know I've messed around with ulimit and with apache2.conf:
<IfModule mpm_prefork_module>
    StartServers          5
    MinSpareServers       5
    MaxSpareServers     100
    MaxClients          150
    MaxRequestsPerChild   0
</IfModule>
More details here (if needed): http://www.linuxquestions.org/questions/showthread.php?p=4577927

I was encountering a similar problem too and for some reason Mike's hint 
(registering an empty/exit function on the SIGPIPE signal) solved the problem for 
me.
The only problem I now have is racking my brains over WHY?! Is this discussed 
somewhere so I can read up on the given solution?

I must correct: the SIGPIPE sighandler does suppress the message but does not 
solve the problem! The command that generates the broken pipe fails.

Ok, so I finally think to have cracked the problem:
This seem to be the equivalent of exec('cmd', &$output) that does not create a 
broken pipe:

$p = proc_open("ls | ls | ls | ls | ls | ls | ls | echo `date` >> foobar | ls",
  array(
    array("pipe","r"),
    array("pipe","w"),
    array("pipe","w")
  ),
  $pipes
);
while (($buffer = fgets($pipes[1])) !== false) $output[] = trim($buffer);
foreach ($pipes as $pipe) fclose($pipe);
if (isset($output)) print_r($output);

There is no broken pipe and the file foobar now contains the output of date.

The above proc_open solution does work (better) for the `ls | ls | ls | ls | ls` 
test case, but not my real-world problem. This is becoming a nasty bug(?) to work 
around.