RFC: Patch policy

RFC: Patch policy

[Robert Schwebel]

Hi Marcin, Richard & OE crew,

At ELC-E, we discussed how we could improve the collaboration between
the cross build systems out there. Although there are big differences
in what kind of tools people use, there is one thing that all cross
build systems have in common: patches for upstream packets.

What people usually do when starting a new packet which doesn't build
out of the box is looking around if oe, buildroot, ptxdist, t2, ... has
a patch, copy it, or, if not, invest more or less time to fix (or hack
around) the upstream issue. Then patches go into the build systems and
bit rot there :-)

I'm wondering if there is interest in more collaboration wrt. bringing
cross development paches upstream; one effort we started after FOSDEM
this year is the send-patches.org project, mainly by creating the
crossdev@send-patches.org mailing list. However, I suppose we should
think about what we could do to increase the activities.

The main concern of the ptxdist maintainers with simply taking patches
from other build systems is patch quality; our rules are:

- patches have to be in the canonical patch format, as known from linux
- proper Signed-off-by: lines
- patches should be made with upstream in mind, i.e. "correct" fixes
  instead of quick hacks; if we really need hacks, they should be
  clearly marked as "not for upstream".
- modify autotool files separately from their autogenerated files

What do you think? Should we try to create a new patch stack which
follows these rules, in order to lower the ammount of duplicate work? At
least for us, this would be very interesting.

Cheers,
Robert (ptxdist team)
-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: RFC: Patch policy

[Leon Woestenberg]

Hello Robert and crews,

On Fri, Oct 30, 2009 at 9:47 PM, Robert Schwebel
<r.schwebel@pengutronix.de> wrote:
> Hi Marcin, Richard & OE crew,
>
> At ELC-E, we discussed how we could improve the collaboration between
> the cross build systems out there. Although there are big differences
> in what kind of tools people use, there is one thing that all cross
> build systems have in common: patches for upstream packets.
>
> - patches have to be in the canonical patch format, as known from linux
> - proper Signed-off-by: lines
> ...
> What do you think? Should we try to create a new patch stack which
> follows these rules, in order to lower the ammount of duplicate work? At
> least for us, this would be very interesting.
>

- Standard naming of the patch (numbered series etc).
- A shared (possibly distributed) patch server with patches against the package?

Regards,
-- 
Leon

Re: RFC: Patch policy

[Robert Schwebel]

On Fri, Oct 30, 2009 at 10:51:02PM +0100, Leon Woestenberg wrote:
> Hello Robert and crews,
> 
> On Fri, Oct 30, 2009 at 9:47 PM, Robert Schwebel
> <r.schwebel@pengutronix.de> wrote:
> > Hi Marcin, Richard & OE crew,
> >
> > At ELC-E, we discussed how we could improve the collaboration between
> > the cross build systems out there. Although there are big differences
> > in what kind of tools people use, there is one thing that all cross
> > build systems have in common: patches for upstream packets.
> >
> > - patches have to be in the canonical patch format, as known from linux
> > - proper Signed-off-by: lines
> > ...
> > What do you think? Should we try to create a new patch stack which
> > follows these rules, in order to lower the ammount of duplicate work? At
> > least for us, this would be very interesting.
> >
> 
> - Standard naming of the patch (numbered series etc).

We use quilt series in ptxdist, i.e. an series file which contains the
names of the patches in order. This has the advantage that you don't
have to break the history by changing numbers when you have to insert a
patch between two others.

> - A shared (possibly distributed) patch server with patches against
> the package?

I could imagine a central patch server for that, as long as the
maintainers have a technique to "qualify" which patches to take for
their respective projects. It could for example be done in a way that
the maintainers add their Signed-off-by, and a script extracts things
automatically then.

rsc
-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: RFC: Patch policy

[Michael 'Mickey' Lauer]

> What do you think? Should we try to create a new patch stack which
> follows these rules, in order to lower the ammount of duplicate work? At
> least for us, this would be very interesting.

Sounds good. Although the number of patches we have been able to land
upstream is quite depressing, I still think it's a worthwhile goal, if
just for reducing duplicated work.

:M: