[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: testing and no release schedule

On Fri, Apr 02, 2004 at 01:03:37PM +0200, Adrian Bunk wrote:
> On Fri, Apr 02, 2004 at 05:23:14PM +1000, Anthony Towns wrote:
> > if you're the RM, because you have to approve _every single upload_,
> > probably for months on end. For potato, for the five or six weeks I was
> > acting RM, that was some 300 changes; since potato the archive's grown
> > by three or four times, and the potato freeze went for something like
> > eight months, not six weeks. And that's not to mention that for potato,
> Who said RM is a lazy job without work?  ;-)

*shrug* If you want to argue that the RM gets to dictate release policy,
that's fine. I've said what I think, and if you believe that RM dicta
are the be-all and end-all, well, we're done here.

> > we thought a hundred RC bugs was almost impossible to manage; and weren't
> > even trying to ensure absolutely everything had its dependencies met,
> > let alone its build-dependencies.
> AFAIR, in potato on i386 all dependencies were met,

AFAICS you recall incorrectly. 


...indicates there were 8 i386 packages in potato when it was stable that
were uninstallable (which only considers dependencies, not recommendations
nor priorities); and 442 packages all up. That compares to woody at the
moment, with 16 such packages on i386, and 225 in total across the same
architectures. There were approximately twice as many packages in total
on each of those architectures (my scripts haven't been generating stats
for the other architectures) in woody compared to potato.

> and there was 
> exactly one Recommends that couldn't be fulfilled. This was far better 
> than the state in woody.

Please don't make claims without both providing the data to back them up,
and making sure they're correct.

> In big commercial project, you typically denote a forth or even half of 
> the development time for testing [1].

We do testing concurrently with development. How many developers aren't
running testing or unstable systems full time? How many users are running
testing or unstable systems in mission critical roles?

> Depending on whether there will be a d-i final, the estimated time
> between the actual freeze and the release will be a few days or a month 
> plus a few days. Is this really enough time for all important bugs to be 
> found and reported?

We've already tested this hypothesis: we had a real freeze of around
a month for woody; followed by a three month period in which woody was
pretty much at the equivalent of zero kelvin. That there wasn't a great
need for changes during that latter period indicates the testing we give
programs in unstable and testing is actually sufficient for our purposes.

It's quite possible it's not sufficient for _your_ purposes, of course,
and you're certainly encouraged to suggest ways Debian could better
fulfill your goals. But you don't get to do that by dismissing the value
of "short freezes". Sorry.


Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

             Linux.conf.au 2004 -- Because we could.
           http://conf.linux.org.au/ -- Jan 12-17, 2004

Attachment: signature.asc
Description: Digital signature

Reply to: