Re: GCC Status (2004-08-09)

["Giovanni Bajo" <giovannibajo at libero dot it>]

Mark Mitchell wrote:

>> Another solution that have been proposed in the past is to alternate
>> Stage 2 and Stage 3 every other week. Of course, there has to be
>> some kind of commitment on actually *working* on bug fixing during
>> the Stage 3 weeks, otherwise this is going to slow down things,
>> instead of helping making mainline stable faster.
>>
>>
> That is exactly my concern.   It also means that we have to make sure
> everyone remembers these switches back-and-forth, and people are going
> to complain when their patch doesn't get submitted/reviewed until
> Monday and now it's temporarily stage 3, and so the patch can't go in for a
> week.

That's right.

I think the main problem is that, at this point, there are many relatively big
patches going in, and only a few bugs fixed every now and then. I hear people
that are delaying bugfixes in their todo list because there is this or that big
change they are preparing. So the situation is going to get worse in the next
weeks as we approach Stage 3, not better. Plus, we got a lot of complaints last
year because Stage 3 was too long.

So I still think that alternating Stage 2 and Stage 3 for a while makes sense,
and enter Stage 3 fully only when the number of regressions is below a certain
level (say, 120 or so). We probably have just to work out a correct way of
doing it, changing every other week is surely too confusing. For instance,
Stage2 for 1 month, then Stage 3 until 30 bugs are fixed, then back? Does this
tackle your concerns, or you think any kind of alternation is going to be
confusing, no matter what?

Giovanni Bajo