[Dev] Rotating build sheriff duties?
Phillip J. Eby
pje at telecommunity.com
Wed Dec 7 11:55:41 PST 2005
At 02:28 PM 12/7/2005 -0500, Mike Taylor wrote:
>What people *need* to do that they are not is simply watch the Tinderbox
>page - when you commit and the tree starts to show failures all devs who
>commited code should be looking at the reason why.
>It really is that simple - the rotating sheriff duty would not be needed
>at all if each dev would monitor the build status page when they commit.
So, the low-hanging fruit would be for us to email the test failures to
developers who committed since the last success of that Tinderbox,
yes? (And in the absence of any commit since last success, we could email
them to the author of the failing line(s) of code, as determined by svn
blame. But that's slightly less low-hanging.) It might also make sense to
have a "builds" mailing list where all test failures would be sent. (As
links to the logs, possibly with highlights, not the actual failures.)
Anyway, to implement that, we'd need to know the subversion revision of the
last successful build on a given Tinderbox, and the subversion revision of
the current build. We'd also need a mapping from developer logins to email
addresses, unless the local accounts are set up to forward to developers'
real emails. And we'd need to parse "svn log", or use the equivalent
Python SVN API calls. I already have some Python code in setuptools that
can scan an SVN checkout to find its effective revision number, so that
part is easy; we just need a place to store the "last successful" one.
If need be, though, I could implement a tool to do all this by
screen-scraping the existing logs, without touching the actual tinderbox
processes at all. The logs already show the svn update commands and the
revisions in question.
By the way, has the tinderbox page moved? I just went to:
and got a 404.
More information about the Dev