fresnel-fourier iter5b-β Phase 6 commit B: config.c — wire object_config->pixelformat

Populate the previously-dead pixelformat field at config.h:46 from
pixelformat_for_profile(profile). The switch at lines 54-88 already
rejects unsupported profiles, so by the time we reach the assignment
at line 98, pixelformat_for_profile returns non-zero.

Commit C reads this field at CreateContext to set the V4L2 OUTPUT
format correctly per profile (the β architectural fix for Bug 2 —
HEVC/VP9/VP8 currently dispatch through the pre-iter5b H264_SLICE
hardcode at surface.c:173 because surface.c has no config_id to look
up the profile).

Signed-off-by: claude-noether <claude-noether@reauktion.de>

src/config.c

@@ -36,6 +36,7 @@
 
 #include <hevc-ctrls.h>
 
+#include "codec.h"
 #include "utils.h"
 #include "v4l2.h"
 
@@ -97,6 +98,16 @@ VAStatus RequestCreateConfig(VADriverContextP context, VAProfile profile,
 
 	config_object->profile = profile;
 	config_object->entrypoint = entrypoint;
+	/*
+	 * iter5b-β: cache the V4L2 OUTPUT-side FOURCC for this profile so
+	 * context.c::RequestCreateContext can read it without re-running
+	 * the profile→pixelformat mapping. Wires up the previously-dead
+	 * pixelformat field at config.h:46. The switch above already
+	 * rejected unsupported profiles via VA_STATUS_ERROR_UNSUPPORTED_PROFILE,
+	 * so pixelformat_for_profile here returns non-zero for every
+	 * profile that reaches this assignment.
+	 */
+	config_object->pixelformat = pixelformat_for_profile(profile);
 	config_object->attributes[0].type = VAConfigAttribRTFormat;
 	config_object->attributes[0].value = VA_RT_FORMAT_YUV420;
 	config_object->attributes_count = 1;