Boost Testing :
From: Beman Dawes (bdawes_at_[hidden])
Date: 2007-10-14 13:47:30
Rene Rivera wrote:
> Beman Dawes wrote:
>> Rene Rivera wrote:
>>> Beman Dawes wrote:
>>>> I noticed your commit "Partial rework of regression scripts for branch
>>>> independent testing."
>>>> Can I do anything to help with this work? I'd really like to get testing
>>>> started on the release branch, and would be happy to do the report
>>>> generation here.
>>> I'm almost done with the changes...
>> Hum... I looked at src/regression.py and realized you've got something
>> different in mind than I was expecting.
> :-) My goal is to make the testing tools independent of the branch one
> is testing. Since it's a rather bad design to have your testing system
> rely on the same code one is testing. I know I can't completely get
> there, because of PJL, but I can get close.
I'm really, really, worried that doing anything more than a minimal
change will cause a delay and reduce reliability while the bugs are
How long are your changes going to take and what is the likelihood they
will introduce new bugs or dependencies?
>> My expectation was that regression.py changes would be limited to being
>> able to pass in the path on the ftp server, currently hard-wired to
>> '/boost-regression'. I'd probably also allow the user to change the ftp
>> server URL, currently hardwired to 'fx.meta-comm.com', but that isn't
>> strictly necessary.
> Should be easy to add in now that I've redone the option handling.
>> Note that I assume that all branches/release testers will do svn
>> updates, at least initially. The tarball is just too unreliable.
> I'm keeping the tarball option in the script for now, but I won't make
> it the default. We'll need to ask the testers to set up svn such that it
> can get around their firewalls. Hence we need to point them to the
> instructions about that in the svn book.
That doesn't seem unreasonable for release testers, particularly since I
don't want them using the tarball if they can possibly avoid it.