282534 - The "Most Frequently Reported Bugs" contains too many obselete bugs

Status: RESOLVED WONTFIX

(bugzilla.mozilla.org :: General, defect)

Description

[Darlene Ford]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8b) Gecko/20050216
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8b) Gecko/20050216

All users reported mozilla bugs are requested to review this page, which is
instended to list the "most frequently reported bugs" in mozilla products.  In
fact, it lists over a thousand bugs, most of which have not been reported in the
last 30 days.  A more concise list, perhaps 300 recent duplicates, would be far
more useful, and user compliance with the request to read the entire list would
improve.  



Reproducible: Always

Steps to Reproduce:
1.Go to
https://bugzilla.mozilla.org/duplicates.cgi?sortby=count&reverse=1&maxrows=1000&changedsince=30
2.  Inverse sort by number of changes in last 30 days.
3.  Note that none of the first 1000 bugs listed have been changed in the last
30 days - unlikely that a bug in the current release is a duplicate of any of
these.  Note also that some bugs have negative changes, which may also be
a bug.

Actual Results:  
I got over 1000 bugs that aren't reported in the latest release after all.  Is
it really reasonable to ask people to read them all before checking in bugs to a
current release?

Expected Results:  
What is the right way to define active bugs?  Should this list the bugs reported
most frequently in the last month?  3 months?  Would a weighted list (e.g.top
100 for metric = # of duplicates reported this month + 0.5*number reported last
month + 0.25 * ... be too computationally intensive?

Comment 1

[Dave Miller [:justdave]]

It's manually triggered, because some bugs that are old still get reported a lot
for some reason, so those particular ones need to stay on the list.  (People
working from old versions, etc).  The report counts anything which is resolved,
but not verified.  A bug which is verified will drop off of the list.

This is a QA issue...  Bugs that are old enough that they shouldn't be on this
list anymore should be getting marked verified so they'll drop off the list.

Comment 2

[Darlene Ford]

No, this still lists bugs marked as "verified" - for instance, 25537 is near the
top of the list 
https://bugzilla.mozilla.org/duplicates.cgi?sortby=count&reverse=1&maxrows=1000&changedsince=30

Could this be configured this way on purpose in case people with older versions
try to report bugs that have already been fixed?  It would make more sense to
run the script you use to check build ID, and redirect anyone whose browser is
old enough that you won't take them seriously to the download page.

Why do some bugs (such as 40934 and 112315) get listed as having negative
changes?  That can't be right...

Comment 3

[Dave Miller [:justdave]]

Hmm, that's a good question.  I think Gerv owns this page.

Comment 4

[Gervase Markham [:gerv]]

The new b.m.o. front page (bug 279823) has better duplicates links on it.

The actual algorithm is not just "if not VERIFIED", it's more complex, and looks
at the resolution too. This is good - we *want* bug 25537 on the mostfreq page.

Negative changes can be right if bugs get un-duped.

Gerv

Comment 5

[Gervase Markham [:gerv]]

Resolving WONTFIX.

Gerv