On Thu, Jun 01, 2017 at 07:25:46PM +0100, Emil Velikov wrote: > On 1 June 2017 at 14:00, Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> wrote: > > Daniel started a crusade a few years back to move control over the > > initialisation and teardown into the driver rather than drm core, for > > greater control and far fewer surprises. Help in that fight by adding > > compiler warnings to the stale functions. > > > > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > > --- > > > > This is not as useful as I hoped. I wanted for gcc to emit a > > compile-time warning when a deprecated field was assigned, however it > > only emits when it is used. So we end up with a lot of noise in drm_core > > and none in the drivers. However, the runtime warning is probably useful > > and a motivator for change? > > > > Should be made WARN_ONCE though > > > I recall suggesting something similar a while back, but Daniel was not > too happy with it. > Hope it goes well this time. The problem is that if we enable it by default people scream regressions! If we enable it behind a Kconfig, Linus screams Kconfig! And no one else will look at it. I'm half-tempted to add cocci-check support to dim to catch these, but that's also an imcomplete solution. I'd love to enforce this with tooling though, one way or another, I just haven't seen something that works well yet :-/ -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
