Re: [PATCH 2/2] Fix t3404 assumption that `wc -l` does not use whitespace.

From: Jeff King <peff@peff.net>
Date: May 15, 2008, 11:20 UTC
Message-ID: <20080515112030.GA12781@sigill.intra.peff.net>
In-Reply-To: <e2b179460805150316n77513037y5409042b01170d4e@mail.gmail.com>

On Thu, May 15, 2008 at 11:16:27AM +0100, Mike Ralphson wrote:

> Which branch(es) would it be most useful on which to have this
> automated build/test cycle?

I would think maint, master, and next, but with next as the least important. I think Junio generally tests maint and master before publishing, but presumably not always next (as there was test breakage in next earlier today).

Show 7 quoted lines

> Although the list of tags might get slightly unwieldy (i.e. the top
> commit will gain a lot of tags if all is well), with a sensible naming
> convention, these tags could be pushed to a central repo (a regularly
> updated clone of git.git) allowing easy visibility of the current
> state of the 'build collective'.
> 
> Something like {intials}_{uname info}_{branch}_KNOWN_{BUILDING|PASSING} ?

I have started tagging my auto-builds as you suggest. It should be easy enough to push to a repo.or.cz repository. Although I'm not sure of the utility of auto-publishing this information. Who is going to look at it?

I had assumed a workflow more like "it passes 99% of the time; in the remaining 1%, the cron job kicks off a message to the owning user, who then investigates and/or writes a bug report to the list."

That implies a little bit of expertise and work from the user owning the build, but:

  - presumably it won't happen very frequently

  - they are probably the only person with the resources to diganose and
    fix, anyway, since they are the ones with access to the platform.

-Peff