On 07/22/2011 11:13 AM, Grant wrote: > Wouldn't a sufficiently large swap (100GB for example) completely > prevent out of memory conditions and the oom-killer?
There's always someone who can pull a corner case out of his hat :) I can't remember the details now, but there was a piece of code in the virtualbox svn repository that triggered a bug in gcc. The code in question caused gcc to loop forever while parsing, using more RAM with every iteration. I watched my 4 gigs of RAM slowly fill up, and then watched nervously as my 4 gigs of swap began to fill up as well. After the swapping started, the machine was almost unusably slow. I typed Ctrl-Z and waited about five minutes until the gcc process finally suspended so I could kill the compile.
