Andrew Lentvorski wrote:
Stewart Stremler wrote:
I'm seeing some hype about automated code-quality tests, and I'm
wondering if we aren't heading towards another annoying fad.
Maybe, but I just consider them to be Lint on Steroids. Fortunately,
most of them are pretty innocuous.
I am starting to do things like set up nagios to monitor services that
should be running and applying tests that should pass, even if
everything isn't working yet. This way I know I am done when there are
no more red alerts on the nagios page. I run nslint on the zone files
every 5 minutes and get email if something breaks. I run cfengine to
enforce that certain config files are a certain way and that certain
symbolic links exist and that certain things are in /etc/hosts or
resolv.conf and that certain binaries are on all of the systems and take
action to fix it or alert me if something unfixable is wrong. I define a
policy and it carries it out. It also restarts processes that should be
running that are not.
Unit testing is along these lines also where you run tests (sometimes
continuously, updating from svn/cvs and running the tests) and are
alerted when something breaks. I am really starting to like the idea of
intentionally creating certain dependencies all throughout the system
(the system being the code, the hosts running the code, then network,
the network supporting infrastructure, etc) that cause alarms to go off
immediately when something breaks and the tests fail but do not
themselves cause any system failures.
I think automated code-quality tests could have a place in such a system
as well as long as you are allowed to dictate the policy because as
evidenced by this thread everyone has a different opinion of what is
quality. But if you were the manager of a big software project wouldn't
you like your local revision control system to be able to run such tests
on code that is checked in and automatically notify the programmer if it
contains things that you the manager would not approve of? Seems like it
would save you time. I don't approve of people making forward dns
entries but no reverse. Without my nslint running every 5 minutes it
would happen often.
With a complete system like this we are looking towards having much more
automated and perhaps even self-healing (meaning something like cfengine
adjusts the configuration back to what the policy dictates when
something goes astray) systems. I tell my coworkers that my goal with
all of this is Skynet. :)
--
Tracy R Reed
http://ultraviolet.org
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
