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

Re: packagekit



On Thu, Dec 11, 2008 at 3:27 PM, Bill Davidsen <davidsen tmr com> wrote:
> This one was a special case, not intended to go out so quickly, but a policy
> change would be welcome, to assure testing of new packages, at least for
> obvious things like permissions, missing requirements, and "won't even
> ____ing boot!" cases.

A policy change? What policy change? You want to take away the ability
for maintainers to rush updates to stable completely? That's going to
be sell compared to the tradeoffs of getting a security update out
quickly. Adding more signoffs to do that sort of thing is going to
just delay important security responses.  dbus was tagged as security
btw...it just wasn't intended for stable.

Are you using updates-testing on any of machines? Are you using
updates-testing on enough enough machines to test hardware dependant
problems which could bite you?
Are you submitting affirmative feedback in bodhi about updates-testing
packages you install?

Even non-booting cases can be quite difficult because they are
invariably hardware dependant. If it boots in my F-9 virtual machine
for example...is it going to boot on your iron? Even if we had
automated testing for "does it still boot" would probably pass because
those tests are going to be done in virtualized systems. Passing the
automated tests may not give us much more info that standard developer
testing.

Any hardware regression can be difficult to catch.  If my laptop
camera still works after an update, is it gonna still work for yours?
Hardware regressions are so very much a pain that can only be
addressed with exceptional testing coverage. Those are the sort of
things that the 'right' people have to be 'lucky' enough to experience
to identify before they can be addressed.

The only way we can prevent users in stable from seeing problems is if
more users volunteer to use updates-testing.  I should generate some
metrics from mirrormanager about the number of people with
updates-testing enabled compared to updates.

-jef


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