Re: CVS debacle

["Jeffrey W. Baker" <jwbaker iname com>]

"Geoff Harrison (mandrake)" wrote:

> brain-dead?  if only you knew that there are an average of over 60+ users of
> each anonymous cvs server at any given time (well - at least dagon is that way
> all of the time - with a load WELL over 40)

I have heavily loaded servers too.  I expect them to be slow, but not to
misplace whole directories (./macros?) or randomly garble source files.
 
> > 2) Ensure that anonymous CVS is in-sync before making announcements.
> 
> it updates once an hour.  can't ask for more than that - rsync would kill ALL
> the machines if it were any more often.

Obvious solution:  Wait until the top of the hour to make your
announcements.

> > 3) Test build gnome before a cvs commit (this seems basic!)
> 
> there's going to be bugs in cvs, pretty much by default.  you use the
> up-to-the-hour code and it's likely it won't always even build - there are bits
> that aren't as often worked on that a lot of folks won't think about testing
> before they commit stuff.

One of the cornerstones of a systematic quality control system is that
the developer must rebuild and run regression tests before committing to
prevent bugs from getting into the trunk of the source repository.  With
gnome, we get a totally backwards QA system: commit all the changes at
once without testing, ignore scores of failing assertions, unleash it on
the public and hope that they will find the bugs.

Personally, I am proud when I send source to testing and they don't send
it back.  I don't expect them to compensate for a slack development
process.

Regards,
Jeffrey