CI patches that are not in series file QA

[Scott Kitterman]

On Tuesday, April 14, 2015 09:25:07 AM Harald Sitter wrote:
> On Mon, Apr 13, 2015 at 9:29 PM, Scott Kitterman <ubuntu at kitterman.com> 
wrote:
> > On Monday, April 13, 2015 06:32:48 PM Harald Sitter wrote:
> >> Another alohas
> >> 
> >> While working on our patch parser today I noticed that there is a
> >> rather unfortunately large amount of patches that get removed from the
> >> series file but not from git (not until someone notices that they are
> >> unused anyway).
> >> So, I just deployed an all new QA measure checking for that and
> >> turning CI builds unstable where the contents of patches/ doesn't add
> >> up with what is listed in patches/series.
> >> 
> >> Random note: since genuine upstream patches do get removed from CI
> >> branches they would potentially still dangle in a landing branch (e.g.
> >> kubuntu_vivid_archvie) if not removed correctly. This situation would
> >> however ultimately auto-resolve when whatever currently used landing
> >> branch gets a CI branch merged (i.e. the upstream branch would then
> >> definitely be removed).
> > 
> > I'm not sure if it's relevant to this or not, but there at least used to
> > be
> > patches that were for Debian only and not Kubuntu that we'd remove from
> > series, but leave it the patches directory to keep the diff smaller/more
> > readable.  Even if this isn't the case for Kf5/Plasma5 yet, I'd guess it
> > will be in the future, so we ought to figure out what to do about it.
> 
> Good point indeed. I think a very simple approach to this would  be to
> simply consider commented out patches in series as fulfilling the
> requirement. As long as the patches are not entirely removed from
> series but only commented out it is still obvious that they are not
> used but at least documented as not being used.
> 
> Unfortunately I fear that simply removing the patch from git (which
> really is the simplest solution) isn't possible as debian would
> eventually merge one of our branches and that would then also remove
> the patch for debian, so that won't be too efficient.

I think commenting out of series is probably the best way to go.  Is there one 
of the policies this can be written down in so we don't forget?

Scott K