On Wed, Mar 2, 2011 at 4:35 PM, Platonides <[email protected]> wrote:
> Rob Lanphier wrote:
>> I'd like to add something to the great suggestions that have already
>> come in.  Since you're already doing daily updates and seeing the
>> breakage, you already know within 50-60 revisions of where the
>> breakage is likely to be.  It may only take 6-7 extra checkouts to
>> narrow it down to a specific revision doing a binary search through
>> the previous day's checkins, even without an informed guess on which
>> file or function is the likely suspect and including extensions.  If
>> you narrow it down to just core (not extensions), that's roughly 20
>> checkins a day, so 4-5 extra checkouts should do the trick.  Any
>> little bit of debugging like this is greatly appreciated!
>
> That would be nice, of course. But we shouldn't demand to debug the
> problem to people which is already doing us a favour by detecting them.
> Just a bug report like: "Foo broke in last 24h"
> "Updating from r12345 to r54321 the foo links are now like bar." would
> be extremely useful, as that bug will be easy to resolve just in time
> (even if it's a revert), as opposed to handling a bug which has been
> there for a long time.

Oh, sure.  It could be that someone like Huib identifies a problem
range, and someone else actually narrows it down to a specific commit.
 I guess I should have explicitly cast the net out wider for the
second part.  The second part doesn't require a deep knowledge of the
codebase, so it's an activity that someone just getting started can do
to help out.

> Those 'fast' bugs are also good candidates for the weekly sprints.

Yup, they certainly are.

Rob

_______________________________________________
Wikitech-l mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikitech-l

Reply via email to