On Wed, 22 Jan 2025, Gustavo Sousa <gustavo.sousa@xxxxxxxxx> wrote: > Quoting Jani Nikula (2025-01-22 11:02:31-03:00) >>On Wed, 22 Jan 2025, Gustavo Sousa <gustavo.sousa@xxxxxxxxx> wrote: >>> Quoting Simona Vetter (2025-01-22 08:11:53-03:00) >>>>On Tue, Jan 21, 2025 at 06:09:25PM -0300, Gustavo Sousa wrote: >>>>> The header drm_print.h uses members of struct drm_device pointers, as >>>>> such, it should include drm_device.h to let the compiler know the full >>>>> type definition. >>>>> >>>>> Without such include, users of drm_print.h that don't explicitly need >>>>> drm_device.h would bump into build errors and be forced to include the >>>>> latter. >>>>> >>>>> Signed-off-by: Gustavo Sousa <gustavo.sousa@xxxxxxxxx> >>>>> --- >>>>> include/drm/drm_print.h | 1 + >>>>> 1 file changed, 1 insertion(+) >>>>> >>>>> diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h >>>>> index f77fe1531cf8..9732f514566d 100644 >>>>> --- a/include/drm/drm_print.h >>>>> +++ b/include/drm/drm_print.h >>>>> @@ -32,6 +32,7 @@ >>>>> #include <linux/dynamic_debug.h> >>>>> >>>>> #include <drm/drm.h> >>>>> +#include <drm/drm_device.h> >>>> >>>>We much prefer just a struct device forward decl to avoid monster headers. >>>>Is that not doable here? >>> >>> I don't think so. This header explicitly uses members of struct >>> drm_device, so the compiler needs to know the full type definition. As >>> an example see the definition of drm_WARN(), it uses (drm)->dev. >> >>I grudgingly agree. I don't think there are actual cases where this >>happens, but I can imagine you could create one. > > It happened to me, and that motivated me to send this patch. > > I had a local patch where I just needed the drm_print.h header, but I > ended up having to include drm_device.h in my .c file. Right. Reviewed-by: Jani Nikula <jani.nikula@xxxxxxxxx> > >> >>>> Worst case I'd convert the drm_info_printer() static inline to a >>>> macro, not sure about the exact rules here if you never deref a >>>> pointer. >> >>The forward declaration is enough for passing pointers around without >>dereferencing. It's the dereferencing in the macros that could fail the >>build if the .c using them doesn't include drm_device.h. >> >>To balance things out, I think we could probably drop drm/drm.h if we >>inlined one use of DRM_NAME to just "drm". >> >> >>BR, >>Jani. >> >> >>>>-Sima >>>> >>>>> >>>>> struct debugfs_regset32; >>>>> struct drm_device; >>>>> -- >>>>> 2.48.1 >>>>> >>>> >>>>-- >>>>Simona Vetter >>>>Software Engineer, Intel Corporation >>>>http://blog.ffwll.ch >> >>-- >>Jani Nikula, Intel -- Jani Nikula, Intel
