[PATCH 1/3] email-test-patchset: automatically obtain a base commit to test against

Keller, Jacob E jacob.e.keller at intel.com
Thu Mar 13 10:36:13 PDT 2014


On Thu, 2014-03-13 at 10:28 +0200, Artem Bityutskiy wrote:
> On Tue, 2014-03-11 at 08:27 -0700, Jacob Keller wrote:
> > In our environment, a queue of patches is used. Often, developers will
> > submit a v2 of a patch already in the queue before the maintainer can
> > "pop" this patch off the queue. This function setup here enables
> > determining if a patch with the same canonical subject is already
> > applied to the tree. A special parameter "base" is used to prevent going
> > back before the "published" upstream branch of the tree.
> > 
> > This enables aiaiai to be smart enough to figure out when a patch should
> > be tested against the tip of queue, or against the parent of a previous
> > version. This relies on new patch versions having identical canonical
> > subject lines, though. (Equivalent except for the [... PATCH ...]
> > prefix(es)
> 
> What do you think would it make sense implement it this way.
> 
> 1. Implement a separate script which tries to check whether a similar
> patch is already applied. And if yes, this script returns the commit id
> which has to be used as the base commit for testing.
> 
> May we we could use term 'hooks' for this? Then this specific hook could
> be called something like 'aiaiai-email-newbase-hook' or something.
> 
> BTW, it may make sense to remove the 'aiaiai' prefix form all
> non-user-facing scripts, like the helpers and e-mail... I would not
> mind.
> 
> 2. Add another config option which will list the hooks which
> aiaiai-email-test-patchset has to execute. So the
> 'aiaiai-email-newbase-hook' gets executed, finds the new base, and
> changes the 'pcfg_branch' value.
> 
> And, IIUC, this should be enough. The patch-set gets tested against than
> new base.
> 
> How does this sound?
> 

That sounds better than what I have done so far. I can take a look at
implementing something like this. I like this too since the inbetween
hook could be written in something different, which may help make it
easier than the crazy eval parsing I have to do in order to work right.

Thanks,
Jake


More information about the aiaiai mailing list