Re: Regression testing [was: A suggestion for the release notes?]


Subject: Re: Regression testing [was: A suggestion for the release notes?]
From: Mike Nordell (tamlin@algonet.se)
Date: Fri Dec 01 2000 - 06:48:17 CST


Sam wrote:

>If this was set up nicely, I could very easily run this as part of the
>nightly build script, which would allow us nightly regression testing.
>Better yet, you can add test runs to Tinderbox, so we could see
>regressions much more quickly.

Now *that* would be a nice touch!

> 3) Someone needs to write the infrastructure. We might want to do
> this just in shell, which would be pretty easy, most likely.

To iterate many documents the shell is appropriate, yes. But there is
currently one problem, wv doesn't restart from scratch when it does get a
new document during a process' lifecycle (Dom, is it something you're on?),
i.e. it behaves differently is started in a "clean" process compared to is
it's just re-run within an already initialized/run process.

Perhaps we need *two* test programs for this. One that tested import of any
document and then exits, and another one that runs its own "shell filename
expansion" an runs the import repeatedly withing one (just once initialized)
process?

> We also might want
>to use something like DejaGNU [1], which would be more work,
> and neccessitate someone (probably me) learning TCL.

Forget it. Not TCL. That would leave use Win32 morons *installing* even more
on our already burdened boxes. Adding TCL to this would leave at least me
out of the loop. This can be solved with shell scripts. You take care of the
*nix ones and I'll deal with the win32 crap.

> And I just tested it, mass conversion is *really* quick. So this
> wouldn't take to much time, even on tinderbox.

Heh, you want a new document that takes a bit longer? VW conversion is a
*very* time-consuming process (so is importing an RTF document).

>But what we can do now would be very good. Consider it a version 1.0
>release goal.

Do you take care of the ReleaseProcedures document?

/Mike - please don't cc



This archive was generated by hypermail 2b25 : Fri Dec 01 2000 - 06:46:29 CST