Patch formatting question: how picky should the robot be?

[Dan Kegel]

So, git generates patches like this:

diff --git a/dlls/ddraw/tests/dsurface.c b/dlls/ddraw/tests/dsurface.c
index da66e2e..4569997 100644
--- a/dlls/ddraw/tests/dsurface.c
+++ b/dlls/ddraw/tests/dsurface.c

and that's the way we like it.

And sometimes people generate patches with cvs, like

http://winehq.org/pipermail/wine-patches/2008-August/059403.html

diff -u -3 -p -r1.57 vga.c
--- dlls/winedos/vga.c	14 Jul 2008 10:57:58 -0000	1.57
+++ dlls/winedos/vga.c	27 Jul 2008 23:13:08 -0000

(That even used to be the way we liked it.)
That's ok, too, isn't it?
If so, I'll make sure my patch robot can handle it.


But what about patches like

http://winehq.org/pipermail/wine-patches/2008-August/059404.html

--- dlls/appwiz.cpl/appwiz.rc	2008-07-26 09:24:56.000000000 +0900
+++ appwiz.rc	2008-07-28 09:48:00.000000000 +0900
...
--- /dev/null	2008-07-28 09:27:06.210506864 +0900
+++ Ko.rc	2008-07-28 09:48:00.000000000 +0900

The patch robot should probably complain about those, right?
- Dan

[Zachary Goldberg]

On Mon, Aug 11, 2008 at 4:51 PM, Dan Kegel <dank(a)kegel.com> wrote:
> So, git generates patches like this:
>
> diff --git a/dlls/ddraw/tests/dsurface.c b/dlls/ddraw/tests/dsurface.c
> index da66e2e..4569997 100644
> --- a/dlls/ddraw/tests/dsurface.c
> +++ b/dlls/ddraw/tests/dsurface.c
>
> and that's the way we like it.
>
> And sometimes people generate patches with cvs, like
>
> http://winehq.org/pipermail/wine-patches/2008-August/059403.html
>
> diff -u -3 -p -r1.57 vga.c
> --- dlls/winedos/vga.c  14 Jul 2008 10:57:58 -0000      1.57
> +++ dlls/winedos/vga.c  27 Jul 2008 23:13:08 -0000
>
> (That even used to be the way we liked it.)
> That's ok, too, isn't it?
> If so, I'll make sure my patch robot can handle it.
>
> But what about patches like
>
> http://winehq.org/pipermail/wine-patches/2008-August/059404.html
>
> --- dlls/appwiz.cpl/appwiz.rc   2008-07-26 09:24:56.000000000 +0900
> +++ appwiz.rc   2008-07-28 09:48:00.000000000 +0900
> ...
> --- /dev/null   2008-07-28 09:27:06.210506864 +0900
> +++ Ko.rc       2008-07-28 09:48:00.000000000 +0900
>
> The patch robot should probably complain about those, right?
> - Dan

Are there factors in this decision other than Alexandre's convenience?

[Dan Kegel]

On Mon, Aug 11, 2008 at 2:27 PM, Zachary Goldberg <zgold550(a)gmail.com> wrote:
> Are there factors in this decision other than Alexandre's convenience?

He's the main voice, but others might feel strongly one way
or the other, too.

[Vitaliy Margolen]

Dan Kegel wrote:
> On Mon, Aug 11, 2008 at 2:27 PM, Zachary Goldberg <zgold550(a)gmail.com> wrote:
> > Are there factors in this decision other than Alexandre's convenience?
>
> He's the main voice, but others might feel strongly one way
> or the other, too.

It's been the rule for a long time when we were using cvs - always diff from 
  the top of the tree.

Detecting git patch is pretty straight forward. With other patches, if you 
can't find referenced file starting from the top of the tree - generate 
warning and drop the patch.

Vitaliy.