Re: Blocker process: tracker bug / whiteboard naming proposal

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

 



On Tue, 2013-01-22 at 11:09 -0600, Bruno Wolff III wrote:
> On Tue, Jan 22, 2013 at 09:42:22 -0700,
>    Chris Murphy <lists@xxxxxxxxxxxxxxxxx> wrote:
> >
> >ProposedBlocker
> >Blocker
> >RejectedBlocker
> >ProposedFreezeException
> >FreezeException
> >RejectedFreezeException
> >
> >And I assume it's still the case each of those has three modifiers: Alpha, Beta, Final.
> 
> I think we can live without prefixes for the whiteboard. There could be cases 
> where a bug is freezeexception for alpha and blocker for beta, but those 
> could be handled by not marking the state for past the next type of 
> release. The simplicity of naming probably gains more than the extra effort 
> needed for a few bugs.

Honestly I have concerns about the last few proposals:

1. (cmurf) "I would drop "accepted". It's a given if it's Blocker or
FreezeException, that it has been agreed upon."

No it isn't, because of the blocker review process. We need to indicate
three states:

Proposed
Accepted
Rejected

We cannot do this with only the Blocks: field, which is why the
whiteboard labels exist. In theory you could tell whether a bug which
blocks the tracker is 'proposed' or 'accepted' by checking the comments,
but that's a big inconvenience for a human to do and impossible for a
tool - such as the blocker tracking webapp - to do reliably. We
absolutely need the extra bit of metadata to indicate this status. A
whiteboard field isn't the perfect way of doing it, but it works.

Keywords are a bit better as you can't make errors entering them, but
actually flags would be the 'best' way to do it in Bugzilla, as they
give us the trinary state explicitly. If we're going to try and address
this within the limits of BZ, we should probably request the RH BZ team
create six flags for us (Blocker/FE for Alpha/Betal/Final), and use
those instead of the Blocks: / Whiteboard: thing. But that involves
relying on the RH BZ team, which past experience has told us can't be
relied upon to get anything at all done remotely quickly. I can contact
them and ask about a timeframe for such a thing, but for now I think
it's worth taking the improvement we can already achieve. Using flags
would also be a rather 'bigger' change than just twiddling with alias
names and whiteboard field names, as it really does involve a
significant adjustment in the mechanism of the process - we'd have to
adjust a lot of docs, transfer all bugs from the F19 blocker bugs to
flags, and do quite a bit of education to ensure people know about the
New Way. It may be more of an F20 thing. Fiddling with aliases and WB
fields is much safer in comparison as the fundamentals of the process
are the same, and the actual bugs don't change. Summary: we really need
to keep the Whiteboard fields for now.

2. (bwolff) "I think we can live without prefixes for the whiteboard.
There could be cases where a bug is freezeexception for alpha and
blocker for beta, but those could be handled by not marking the state
for past the next type of release. The simplicity of naming probably
gains more than the extra effort needed for a few bugs."

Assuming you mean 'suffixes' not 'prefixes' - so your proposal is just
to use Accepted and Rejected - then again, I don't like that. "There
could be cases" where a bug has multiple states is putting it much too
weakly - there are such cases, a lot of such cases, it's something we do
all the time. We can't just handwave it away. I don't think the
'simplicity' of Accepted vs. AcceptedBlocker is worth that at all. In
fact, a whiteboard field which just says 'Accepted' is probably more
confusing than one which says 'AcceptedBlocker', if you don't know the
process.

3. (kparal) "AcceptedBlocker
RejectedBlocker
AcceptedFreezeException
RejectedFreezeException

The latter is a bit long though, any other proposals?"

This one's actually fine, but if you're worried about length, the
alternative is to just go with what tflink initially suggested, adjusted
for the new name:

AcceptedBlocker
RejectedBlocker
AcceptedFE
RejectedFE

Shorter, less self-explanatory. Looking at them side-by-side, I'm
actually happier with the long version. It's not like we're being
charged by the character, and 'AcceptedFreezeException' is quite a lot
easier to figure out the meaning of if you don't already know about this
process than 'AcceptedFE'.

(as a side note on all the camel casing - it's worth noting that BZ acts
like Windows in this regard, visually it displays aliases, WB fields etc
as cased, but internally it treats them as case-insensitive. If you mark
a bug as blocking 'acceptedblocker' instead of 'AcceptedBlocker' it'll
work just fine. If you search for 'rejectedfreezeEXCEPTION' instead of
'RejectedFreezeException' then again, it'll work just fine. So you don't
have to be totally precise with the camel casing.)

I think we've got this proposal to a really good place, though, and we
all agree that it's a big improvement. In the interests of keeping the
momentum going, unless anyone has any really big objections, I plan to
put the change into 'production' today. We have a bit more time to argue
about the whiteboard field names as we don't start using them until we
start having blocker meetings, but here's what I plan to do today:

1. Give both the non-versioned aliases (AlphaBlocker, BetaBlocker,
AlphaFreezeException etc) and new-format versioned aliases
(F19AlphaBlocker, F19BetaBlocker etc) to the F19 bugs, along with the
'old format' aliases - the F19 bugs will have three aliases each for now

2. Create the F20 bugs (this is supposed to happen now as part of
housekeeping anyway) with the new-format versioned aliases
(F20AlphaBlocker, F20AlphaFreezeException etc)

3. Add new-format versioned aliases to the bugs for all previous
releases (F18AlphaBlocker, F15FinalFreezeException etc)

4. Adjust the wiki docs (SOPs especially, and anywhere else I remember)
for the new aliases and whiteboard fields - I'll just use kparal's WB
field names for now, this is simple to change if we pick a different way
of doing it

5. Ask the RH BZ dev team whether we could get flags created in a
reasonable timeframe if we wanted to go that way

All of these changes are quick, simple, non-invasive, non-destructive
and entirely changeable and reversible, so I don't think there's a
problem with Just Doing It. Again, if anyone has a big concern, please
raise it in the next hour or so, or else I'll start cranking. Thanks
much to everyone who chipped in and improved the initial proposal, I
think this was a great example of FUDCon providing some momentum but not
excluding people who weren't there from the process - the current
proposal is clearly better than the one we initially came up with!

If people have opinions on the idea of going the flag route, please do
toss 'em in. It's worth noting that Tim's long-term idea for fixing the
limitations of indicating blocker state with Bugzilla metadata is simply
to do it in the blocker tracking webapp, which has quite a lot of
advantages and is probably the best long-term option. So there's a
question whether the benefits of using flags are worthwhile if it'll
only be for a few releases until the blocker tracking webapp is able to
do it.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net

-- 
test mailing list
test@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe:
https://admin.fedoraproject.org/mailman/listinfo/test



[Index of Archives]     [Fedora Desktop]     [Fedora SELinux]     [Photo Sharing]     [Yosemite Forum]     [KDE Users]

  Powered by Linux