[opensolaris-summit] [ogb-discuss] My comments (very subjective) on proposed Summit topics

[Garrett D'Amore]

Alan DuBoff wrote:
> On Tue, 25 Sep 2007, Garrett D'Amore wrote:
>
>> There is another real problem here.  I think the only people who are 
>> qualified to "fully" evaluate the bug database are those who are also 
>> Sun employees.  Why?  Because only they really know what the current 
>> bugster contains, and what features are needed, and what aren't.
>
> I just chopped out a bunch of folks on the cc list.
>
> But I wanted to ask, what's the difference anyway? I have seen people 
> complain that they can't read the entire bugs on the OpenSolaris bug 
> database. Ok, even if you could read them internally, there are cases 
> where you still need to look at the actual files to fully understand 
> what is being changed.
>
> Hypothetically, let's say you created a process that made it so 
> difficult for engineers to do their putback, that they started to slip 
> changes into another change request to get their changes back?

That would be an error.  IF the process were so heavy weight, then 
something is wrong.

I've slipped many changes into a single putback ... *but* I try to make 
sure I have only one CR.  (The exception being bugs against a new driver 
or other software, which I haven't integrated.  In that case there is 
only one CR, which is the integration of the driver in the first 
place... in that case it makes sense to try to make the first 
integration as bug-free as you can.)

And, FWIW, I think a lot of the RTI advocates would catch this.  Most of 
the advocates I've used in the past have acted as another layer of code 
review... even though that isn't strictly part of their RTI duties.

>
> This would never happen at Sun, right?
>
> My point is that even if you could read the bug database from outside 
> SWAN, in it's whole, it still wouldn't ensure that you could know 
> exactly what a putback did unless you really looked at every single 
> file and evaluated how they relate to the change request anyway. My 
> $0.02.

You're misunderstanding the problem.

People use the CR database for a lot more than knowing what was in a 
putback.  It contains customer call records, sample code, attachments, 
etc.  A lot of those use cases happen internally to Sun.  I'm not sure I 
even know what they all are.  Someone should involve, for example, RPE 
and CTE in such an evaluation... they are amongst the heaviest bugster 
users.

>
> For the record, I'm not claiming this goes on willy nilly, just that 
> in any large set of sources where you have massive amounts of changes 
> happening on a daily basis, there is a higher chance that engineers 
> will integrate more changes to a CR than logical. Maybe it's a part of 
> common habit.

I don't think so.  Certainly I don't see that in the kernel.  Codereview 
and RTI is supposed to protect against it.

    -- Garrett
>
> -- 
>
> Alan DuBoff - Solaris x86 IHV/OEM Group