The linked PDF document was published by Stanford University's Computer Systems Lab and it discusses automatically detected compiler errors found in 21 snaps shots of the Linux kernel over seven years and baselines these against the OpenBSD kernel.

I know what you are thinking, "This type of error detection is limited and not as useful.", "All Linux bugs are fixed right away, especially source level bugs.", "OpenBSD has less bugs because they work round the clock on pre-emptive bug killing, so this is just an attempt at making Linux look bad." So why would you want to read this document?

First off it answers a few questions about where bugs are likely to occur, if bugs cluster, how long bugs live, and what causes these bugs. Additionally, the following assessments were made that should be of interest:

Average Linux bugs live 1.8 years. (though this may be higher as many bugs are still alive, some as long as seven years.)
OpenBSD showed more errors than Linux in every audit system used.

Also it is discussed that Linux's error rate has decreased over the seven years, which indicates the system is making progress. This is a very important and very unimportant point at the same time from a security stand point. The monolithic architecture creates an instance where ever exception at this level could lead to a full compromise, so this type of hardening is good. However this study fails to cover high level design, so even if the code were completely free of source code bugs security issues could/would still exist.

Although this document is from 2001, there is no reason to believe it's findings have been made obsolete in that time.

Hope you all enjoy this document.

EDITED: AO felt this document was too large, so I have linked to it here:
~download~

catch

PS. Edited AGAIN cause I typoed the path.