On Fri, Mar 13, 2009 at 4:38 PM, Peter Keller <
psilord@...> wrote:
> Out of curiosity, how do you build and test this software? Do you have
> nightly
> regression tests and on how many architectures? Or is it more sporadic,
> like a few people get together and pound on it a bit just before a release?
It's all much less formal than that. The long and short of it is that the
developers run the tests on every new commit, and then we assume it's good
until we hear otherwise =). We have a fair number of people running the
latest SVN version at any given time, and I believe we have a good number of
platforms represented, so anyone who finds a problem in SVN reports it in
our bug tracker and downgrades to an earlier version until we fix it (or
sometimes fixes it themselves). Then we just try to fix any critical bugs
before making a release. If a major bug does get into a release, then again
people can downgrade or use an SVN version. It sounds pretty helter-skelter,
but it's worked well for us.
David
