[Libreoffice-qa] After EOL has been reached...

[Robinson Tryon]

On Tue, Apr 23, 2013 at 12:13 PM, Rainer Bielefeld
<LibreOffice at bielefeldundbuss.de> wrote:
>
> not good! We will not create complicated wrong rules for things what are
> easy to understand.

My "rule" was 1 short sentence long. Compared to the current morass of
QA documentation, I don't think it's appropriate to call that
"complicated."

> And whether a Version still is useful for indicating
> where in the history a (newly reported) bug appeared has nothing to do with
> the date of the last build. So already the subject of this thread is
> misleading for this issue.

Well, now we're talking about a slightly different issue. I agree that
it's important to know when a bug first appeared in our builds, but
that's a QA/triage step, not necessarily something that we need to
front-load to our users.

> LibO-Bugzilla admins will switch old Versions to "inactive" when they see
> that no more (or very few) bugs appear newly in BZ with that version. That's
> the only indicator that counts, that already is in progress, no reason for
> concerns.

I think that there are actually two separate issues here:

1) What versions of LO are users *actively* using and against which we
are seeing bugs reported?
2) What is the earliest version of LO in which we reproduce a bug?

>
> It seems that 3.4 will reach a status what sill allow to switch inactive the
> non-release versions (see statistics for last half year [1]) soon.

Of the 4 bugs you list for 3.4,
1 was a Bugzilla test by Joel (irrelevant)
1 was an AOO bug (!) mistakenly reported in our bugtracker (also irrelevant)
1 was reported against 4.0, and traced back to 3.4.5 by Rainer (so not
reported against 3.4)
1 was reported by a LO developer with 60+ commits to his name (i.e.
largely pre-triaged)

tl;dr: In the last 6 months no ordinary users have reported a bug against 3.4.

Here's a fun Question: Does anyone know off the top of their head how
many 3.4.x versions we have listed in FDO? I just counted, and it's a
whopping 22. All for a release that we consider EOL'd over a year ago!

To quote Sweet Georgia Brown, "Ain't nobody got time for that!"

>
> I would appreciate a stop of this rather useless discussion.

Question: Is the current process documented anywhere? Can we improve
the bugtracker? If no, then why ask us to stop the discussion?

If we can't come to consensus about de-listing the versions from FDO,
how about we at least remove them from the BSA? That way our primary
interface for end-users would be de-cluttered (reduce 22 entries to 1
for 3.4, 14 down to 1 for 3.3, etc..), and we'd still have all 4
glorious 3.3.0 RC's listed in Bugzilla.

On a related note, Joel just added a new item for our next meeting
regarding our policy on bibisect and updating version #'s. I think
that it it's very instructive to know when a bug was introduced into
the code, however bibisect gives us a finer granularity than even our
version picker can provide. In that vein, I wonder if our use of the
version picker is a little myopic -- perhaps there's a better way for
us to keep track of which particular builds are affected by a
particular bug, especially as a bug might appear in one build,
disappear in a subsequent build, and reappear later still[1].

Cheers,
--R

[1] See Joel's notes:
https://wiki.documentfoundation.org/QA/Meetings/2013/May_3#NEW_ITEM:_Clarify_Bibisect_.2B_Version_.23_.28Joel.29