AllHands/BugzillaMetrics

Last year, dmose asked for various metrics with respect to outstanding patches in Bugzilla; here are the results obtained in mid-December 2005.

How many [patches] are there with what distributions of age / patch size / module?

• How many outstanding patches?

SELECT count(attach_id) as count FROM attachments INNER JOIN bugs ON 
attachments.bug_id = bugs.bug_id WHERE ispatch = 1 AND isobsolete = 0 AND 
resolution = "";

7713

• Distribution of age?

SELECT (((YEAR(attachments.creation_ts) * 12) + MONTH(attachments.creation_ts)) 
/ 12) AS period, count(bugs.bug_id) AS count FROM attachments INNER JOIN bugs 
ON attachments.bug_id = bugs.bug_id WHERE ispatch = 1 AND isobsolete = 0 AND 
resolution = "" GROUP BY period;

• Distribution of size?

SELECT TRUNCATE(LENGTH(thedata) / 1000, 0) AS sizek, count(attach_id) AS count 
FROM attachments INNER JOIN bugs ON attachments.bug_id = bugs.bug_id WHERE 
ispatch = 1 AND isobsolete = 0 AND resolution = "" GROUP BY sizek;
(b.m.o. version; it would be different on a tip install)

1k divisions:

5k divisions:

• Distribution of module?

SELECT components.name, count(attach_id) AS count FROM attachments INNER JOIN 
bugs ON attachments.bug_id = bugs.bug_id INNER JOIN components ON 
bugs.component_id = components.id WHERE ispatch = 1 AND isobsolete = 0 AND 
resolution = "" GROUP BY component_id ORDER BY count DESC;

• The review page shows the date that the patch was posted to bugzilla, not the date that review was requested. For how many patches do these dates significantly differ (eg by more than two weeks)?

SELECT count(*) from attachments INNER JOIN bugs ON attachments.bug_id = 
bugs.bug_id INNER JOIN flags on attachments.attach_id = flags.attach_id INNER 
JOIN flagtypes ON flags.type_id = flagtypes.id WHERE flagtypes.name = 'review' 
AND ispatch = 1 AND isobsolete = 0 AND resolution = "" AND  
(TO_DAYS(flags.creation_date) - TO_DAYS(attachments.creation_ts) > 14);

362 (4.7%)

• How man/what percentage of patches have r? set without an email target?

SELECT bugs.bug_id from attachments INNER JOIN bugs ON attachments.bug_id =  
bugs.bug_id INNER JOIN flags on attachments.attach_id = flags.attach_id INNER  
JOIN flagtypes ON flags.type_id = flagtypes.id WHERE flagtypes.name = 'review'
AND flags.status = "?" AND requestee_id IS NULL AND ispatch = 1 AND isobsolete
= 0 AND resolution = "";

71 (0.9%)

• How many/what percentage of patches have no apparent communication from the reviewer that the review will not be immediately forthcoming?

Can't tell automatically.

• Of patches that do have such communication, how many/what percentage is incorrect (ie the reviewer said "I'll review next week", but that didn't happen)?

Can't tell automatically.

• What percentage have already gone through an iteration of the review process? Of those that have, how many / what percentage have already been reviewed by the same person on whom review is currently waiting?

Can't tell automatically.

• How many/what percentage have review requests set against people who are no longer active in the project?

We have to define "no longer active" as "no logincookies entry" (i.e. not logged in in the past 30 days), because Bugzilla keeps no other information on this question.

SELECT profiles.login_name AS name, count(flags.requestee_id) AS reviews FROM 
profiles LEFT JOIN logincookies ON profiles.userid = logincookies.userid INNER 
JOIN flags ON flags.requestee_id = profiles.userid INNER JOIN flagtypes ON 
flags.type_id = flagtypes.id INNER JOIN attachments ON attachments.attach_id = 
flags.attach_id INNER JOIN bugs ON attachments.bug_id = bugs.bug_id WHERE 
logincookies.userid IS NULL AND flagtypes.name = 'review' AND 
attachments.ispatch = 1 AND attachments.isobsolete = 0 AND bugs.resolution = "" 
GROUP BY name ORDER BY reviews DESC;

• How many such inactive review requestees are there?

44

• How many/what percentage are large enough and old enough that they're likely to have bitrotted and no longer be applicable to the current trunk?

See above: the total count is 77, which is not really significant compared to the total of 7713.

• How many/what percentage include complaints about the lack of review?

Can't tell automatically.