Archive for July, 2008

Bugbusting: good news and bad news

Thursday, July 3rd, 2008

There is good news, and bad news, on the bugbusting front.

I think we have made good progress on the state of the PR database recently. In particular, with the help of our recent volunteers attractedby the Bugathons, we have been able to filter the backlog, identifying the majority of the PRs by what manpage they are correlated with (for kern/ PRs, primarily sections 3, 4, and 8 via ‘tags’; for bin/ PRs, primarily sections 1 and 8). There are now prototypes of HTML reports that reports corresponding to each of these. (They are not yet interactive; they run once per day on freefall). Also, all the PRs with patches are now tagged by ‘[patch]‘, and there is an HTML report for those, too. There are over a thousand of these, representing a significant percentage of the database.

In addition, we have added pages to list a highly-filtered set of a few dozen PRs that we think are ready-to-go, and another set of PRs that are the most-commonly-seen set.

Further, by request, we have added some scripts on freefall to allow a committer to pick a random PR to look at.

(I consider all of this work to be effective prototypes of pages that we would need to customize for any PR system. I have seen nothing yet that supports exactly what FreeBSD needs, out-of-the-box.)

(fwiw, most of this has been by suggestion, from either our new bugbusters, or src committers.)

Finally, many of the “I do not know how to make xyz work” PRs have been either closed, or are being handled by the new volunteers.

I have advertised all of this work on the wiki (including the relevant hrefs) on http://wiki.freebsd.org/BugBusting/Resources, and by announcements on the freebsd-bugbusters IRC and mailing lists, as well as selected cross-posting to other mailing lists.

So, what’s the problem?

Well …

This has resulted in only a handful of src commits.

Apparently I haven’t figured out how to get the src committers sufficiently interested to investigate these pages. I have gotten very little emailed feedback, and few committers showed up for the last bugathon.

So, the challenge is to keep our bugbusters interested in doing the triage work. IMHO, to keep the momentum that has been built up, we need to start getting some commits in.

Unfortunately, I do not seem to have the right ideas in place to make that happen. This is very frustrating.

I would be interested in getting suggestions on where to go from here. I am clearly missing something.