On Thu, Jul 13, 2023 at 12:41:15PM +0200, Thomas Zimmermann wrote:
> Hi Uwe
>
> Am 13.07.23 um 12:22 schrieb Uwe Kleine-König:
> [...]
> >
> > If that helps you: I plan to tackle these in a followup. I agree that's
> > ugly, but fixing these in the same series is too much and if we agree
> > that
> >
> > dev = crtc->dev;
> >
> > is ugly, there is then an intermediate state that either used
> >
> > drm = crtc->dev;
> >
> > or
> >
> > dev = crtc->drm;
> >
> > to finaly reach
> >
> > drm = crtc->drm;
> >
> > If you see a better way, please let me know. I don't see any.
>
> IF this renaming gets accepted, you could make a single patch
> that adds a union with both names to all affected DRM data structures.
>
> struct drm_foo {
> union {
> struct drm_device *drm; // use this
> struct drm_device *dev; // don't use this
> };
> }
>
> That patch prepares all of DRM for the renaming. And then you can go through
> drivers one-by-one with a patch or a patchset for each to do all the
> renaming there. Finally you send another DRM-wide patch to remove that
> union.
>
> That would split the changes into smaller pieces without ugly intermediate
> state. It also eases the problem for backporters and stable maintainers that
> have to deal with the name changes.
>
> IIRC you already did something similar for remove or probe callbacks?
Did you look at patch #1 of the initial series? That's exactly what I
did there. People argued to make a single patch and not use an anonymous
union though ...
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | https://www.pengutronix.de/ |
Attachment:
signature.asc
Description: PGP signature
