increase recommended php memory_limit to 384M

Just to update, before I deleted #34's VPS, I tried it for a while on 285M and that still wasn't enough. #32 seems to be happy with 384M.

Turning this into a task to increase the PIFR's self test recommended setting from 256M to 384M.

I'll see if I can setup xdebug to tell me somehow.

xdebug makes a new xdebug_peak_memory_usage() function available at runtime, so there's a way.

Maybe I can just put that at the bottom of index.php, have it insert the value in a file or db table, and return with the results. Let's see what happens.

I'm testing locally, not through PIFR. Here's how I'm testing:

pecl install xdebug

Create a table to store the memory numbers:

CREATE TABLE IF NOT EXISTS `xdebug` (
  `memory_usage` int(11) NOT NULL,
  `peak_memory_usage` int(11) NOT NULL,
  `time_index` double default NULL
) ENGINE=MyISAM DEFAULT CHARSET=latin1;

Insert this on the bottom of index.php:

db_query("INSERT INTO xdebug (memory_usage, peak_memory_usage, time_index) VALUES (%d, %d, %f)", xdebug_memory_usage(), xdebug_peak_memory_usage(), xdebug_time_index());

Activate simpletest and run them all. It gets to the openid tests locally, then I get

Fatal error: Call to undefined function bccomp() in /Users/davidnorman/Sites/drupalhead/modules/openid/openid.inc on line 313

Before then, the largest peak memory consumption value is 126.72238159 MB taking 95.77 seconds on my 2.4ghz Core duo, based on the result from:

SELECT memory_usage /1024 /1024, peak_memory_usage /1024 /1024, time_index
FROM xdebug
WHERE 1
ORDER BY peak_memory_usage DESC
LIMIT 1

Obviously, I'd have to try on a different, un-buggy computer to get the entire test, but that won't happen soon, which is why I put my methodology above. The above also doesn't tell which test is running to know which one is the biggest hog. Maybe there's something in the $GLOBAL environment that would help clue us in on that?

Something is taking more memory. I used to be able to run concurrency 4 with 1GB ram... now it's swapping up to another 1GB to disk. Fields in core? That's the biggest recent thing...

Nevermind. There seems to have been improvements since I filed this.