[qmtest] getting started

[John Schmitt]

> -----Original Message-----
> From: Mark Mitchell [mailto:mark at codesourcery.com]
> Sent: Monday, February 17, 2003 9:57 AM
> Subject: RE: [qmtest] getting started
> [...]
> The menu problem you were having is a known problem -- sort of.  We
have
> seen this in one other case; a user could not get the menus to work.
> Interestingly, logging in on the exact same machine as another user
worked
> fine; it was something -- but we don't know what -- to do with that
> particular user's account.  Unfortunately, we've never been able to
figure
> out what caused that -- but it is almost certainly a Mozilla bug of
some
> kind.

I assume then that it is not perfectly reproducible.  Perhaps, the
Mozilla team could benefit from using QMTest and the pexpect module. ;-)
http://pexpect.sourceforge.net/
http://pexpect.sourceforge.net/doc/pexpect.html
http://mail.python.org/pipermail/python-dev/2002-November/030320.html 

> 
> I apologize for the problems that you had.

No need to apologize at all.

I still have the job of deciding how we're going to do testing and I
have to decide whether to use DejaGnu/Expect, write my own, or use
QMTest.  I'm almost the only real Python advocate here and there are
some expect/tcl advocates here which might carry some weight even though
the implementation languages shouldn't be the deciding factors in this
decision.  

Are there any docs or discussions in the mail archives that discuss
real-world QMTest deployment?  I'm thinking about things like:
- kicking off the nightly build
  -- does QMTest kick it off or does make/cron
     kick off QMTest?
- policy/procedure for maintaining tests
  -- do developers check them in or add them via the
     gui/browser?
  -- if they're checked in, do you make a resource a
     prerequisite install them to a useful place?
- automatically create bugzilla tickets for
  new/unexpected failures?
- when reporting a failure, how can I specify which
  tag should be used to check out the source when
  attempting to reproduce the problem
- how is the test database propogated to the test machines?
  -- share them over nfs?
- are there any database classes that stores tests as
  plain text files?

Doing a cursory scan (by thread) of the mail archives didn't answer
these questions.  Thanks for all the feedback.

John