[Mesa-dev] [PATCH] intel: flush fake front buffer if server is about to destroy it.

[Paul Berry]

Fixes piglit test "spec/!OpenGL 1.0/gl-1.0-front-invalidate-back"
---
This patch depends on "[PATCH 2/2] intel: flush fake front buffer more
robustly.", which was sent to mesa-dev yesterday and is still awaiting
review.

 src/mesa/drivers/dri/intel/intel_context.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 0afd417..4508e49 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -1079,6 +1079,15 @@ intel_query_dri2_buffers(struct intel_context *intel,
 
       attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
       attachments[i++] = intel_bits_per_pixel(front_rb);
+   } else if (front_rb && intel->front_buffer_dirty) {
+      /* We have pending front buffer rendering, but we aren't querying for a
+       * front buffer.  If the front buffer we have is a fake front buffer,
+       * the X server is going to throw it away when it processes the query.
+       * So before doing the query, make sure all the pending drawing has
+       * landed in the real front buffer.
+       */
+      intel_flush(&intel->ctx);
+      intel_flush_front(&intel->ctx);
    }
 
    if (back_rb) {
-- 
1.8.3