On 06/13/2013 03:46 PM, John Arthorne
wrote:
> From: Denis Roy
<denis.roy@xxxxxxxxxxx>
>
> On which day does "last minute" begin, as we are 13 days
away from
> the release? As a casual observer I'm puzzled by the
"test
early,
> test often, fix nothing" mantra. Does today's RC somehow
become
> null and void if a respin is made and serious problems
arise from
it?
CDO is a +2 project, so their RC4 deadline was
Tuesday.
So this is two days after their deadline, which is definitely
"last
minute". My impression is that the bug is serious and we
should fix
it in this case, but I think it was quite reasonable for David
to request
that we follow the process to make sure that all projects are
informed
and have an opportunity to give their input.
This particular case was a perfect example of
why
we need to be cautious and have a process for fixing even
serious problems
at this stage. In this situation a fix for a "serious" problem
two days ago introduced a much worse regression that has now
become a blocker.
This is no reflection on CDO - every project I have worked on
has had cases
like this where a heavily reviewed and tested last minute fix
has introduced
a bug more serious than the one we were attempting to fix.
Sometimes shipping
a known bug is better than risking introducing a worse bug. As
I said in
this case I gave my +1 but sometimes "fix nothing" is the
right
answer.
I naively thought we could spin a RC4a candidate, test it and use it
if it's better, discard it and use current RC4 if it's not.
Denis
|