On 28 May 2015 at 00:56, Joonyoung Shim <jy0922.shim at samsung.com> wrote:
> The build error is introduced by commit fde496917682 ("Add device
> enumeration interface (v4)") if don't enable udev. Can solve as check
> UDEV dependency.
>
> CC libdrm_la-xf86drm.lo
> xf86drm.c:66:21: fatal error: libudev.h: No such file or directory
> #include "libudev.h"
> ^
> compilation terminated.
> make[2]: *** [libdrm_la-xf86drm.lo] Error 1
>
> Signed-off-by: Joonyoung Shim <jy0922.shim at samsung.com>
> ---
> xf86drm.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/xf86drm.c b/xf86drm.c
> index b5a174b..4a31019 100644
> --- a/xf86drm.c
> +++ b/xf86drm.c
> @@ -63,7 +63,9 @@
>
> #include "xf86drm.h"
> #include "libdrm_macros.h"
> +#if defined(UDEV)
The macro you're looking for (here and below) is #if HAVE_LIBUDEV.
> #include "libudev.h"
> +#endif
>
> #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) ||
> defined(__DragonFly__)
> #define DRM_MAJOR 145
> @@ -2819,6 +2821,7 @@ char *drmGetRenderDeviceNameFromFd(int fd)
> return drmGetMinorNameForFD(fd, DRM_NODE_RENDER);
> }
>
> +#if defined(UDEV)
> /**
> * Enumerate the GPU devices on the system
> *
> @@ -2917,3 +2920,4 @@ int drmGetPciDevices(drmPciDevicePtr devSet, uint16_t
> vendorId)
>
> return drmDevCount;
> }
> +#endif
The above change will lead to a broken library as the symbol will be
missing. Just returning -ENOSYS when libudev is missing sounds
reasonable imho. Although I'm more inclined to reverting the patch,
unless we get an actual user and an implementation that does not pull
the dependency of libudev.
Thanks
Emil
