On Friday, April 4, 2014, Brian Wolff bawolff@gmail.com wrote:
Best practise is for author to abandon them if they are not interested any more. However very few people actually do that. If you want to exclude them, -1 is a good criteria (-1 patchsets are not waiting on review, so even if there is still interest, its not in queue).
Makes sense. Let's start excluding open changesets with a last update older than 30 days and -1.
https://bugzilla.wikimedia.org/show_bug.cgi?id=63533
Some [wip] patches are probably also outliers in the data set and should be
excluded.
Can you give examples of this type of changesets? Are there many of these?
For what is worth, in Bugzilla metrics we are able to ignore reports using the "tracking" keyword.
If you want to know median time to review, i would suggest only counting
merged patches, since open patches have an unknown time to review (not sure if this is what you are already doing).
So far the focus has been put into highlighting the areas that need attention, as opposed of the performance measured by the work completed.
For what is worth, there is a "review time in days" graph at http://korma.wmflabs.org/browser/scr.html, but I haven't looked at it, and I'm not sure what is being counted. Perhaps average time instead of median, and therefore getting distorted by patches that get reviewed between team members in a few hours?
In any case, the disparity of age between resolved and unresolved changesets might show that there is a clear stream of fast work, while there is also a pocket of open changes languishing. Let's look at the numbers again when the rulo of ignoring old -1s is applied.