Testing?

[Eric Rostetter]

Quoting Benjamin Smith <lists at benjamindsmith.com>:

> I've been using "testing" on my yum.conf for a FC1 system.

Great!

> See, the system doesn't do much. It plays MP3s 24x7 a la mpg321 (picks songs
> at random from my considerable archives) and it backs up systems via rsync.

It's actually doing a lot then.
 
> It's hard for me to say anything other than "that recent glibc library
> installed OK" since I've not done anything else with it, other than see it
> appear in yum a while back.

The hard part is you have to track what was installed, so you know what
to report as working.  It doesn't do much good to say "all the testing
packages I've installed work okay" if we don't know which packages those
include and which it doesn't include.

If you track the installed versions, then it is of value to us.

> Does this information provide any actual value? Is there some testing harness
> availabe somewhere so I can know "yep" or "nope" package foo works or
> doesn't?

Yes, it is of value.  You system obviously is running the kernel.  So if
you track that you installed a new kernel, and assume you then reboot to
that new kernel, and your machine runs fine for several days, then you have
in effect QA'd that the kernel is stable in your setup.

Also, as you say, if you installed the glibc library, and nothing breaks,
then again you have QA'd that package fairly well (since lots of what you
are doing depends on glibc).

Even if you are not actively using a package, the fact that it installed
cleanly and didn't break anything (or the fact that it didn't install
cleanly, or did break something) is very good information from a QA point
of view.

So, yes, if you track what you install, and report back those packages
and versions and your experience with them, then you are doing QA on them.
Maybe not the best QA possible, but few due that kind of testing (stress
testing, fuzz input testing, exploit testing, etc).  For most people the
testing is 3 questions: does it install, does it work, does anything break?
You can answer those for many packages doing no more than what you do now.

> -Ben
> --
> "The best way to predict the future is to invent it."
> - XEROX PARC slogan, circa 1978

-- 
Eric Rostetter