Merge branch 'mesa_7_7_branch'

Conflicts:
	src/mesa/drivers/dri/r600/r700_assembler.c
	src/mesa/main/version.h

17 files changed, 185 insertions, 69 deletions

diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
index fe2db04d8fa..3eb0e0c57cb 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
@@ -560,9 +560,9 @@ emit_instruction(
       if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ||
           IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ||
           IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z )) {
-         LLVMValueRef *p_floor_log2;
-         LLVMValueRef *p_exp;
-         LLVMValueRef *p_log2;
+         LLVMValueRef *p_floor_log2 = NULL;
+         LLVMValueRef *p_exp = NULL;
+         LLVMValueRef *p_log2 = NULL;
 
          src0 = emit_fetch( bld, inst, 0, CHAN_X );
          src0 = lp_build_abs( &bld->base, src0 );
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index 63ae5c2766a..9c9fc6f64b3 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -36,4 +36,4 @@ include ../../Makefile.template
 .PHONY : $(COMPILER_ARCHIVE)
 
 $(COMPILER_ARCHIVE):
-	cd $(TOP)/src/mesa/drivers/dri/r300/compiler; make
+	$(MAKE) -C $(TOP)/src/mesa/drivers/dri/r300/compiler
diff --git a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
index 6bf3183ff54..56070a1ba10 100644
--- a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
+++ b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
@@ -46,6 +46,7 @@
 #define DRM_VMW_FIFO_DEBUG           11
 #define DRM_VMW_FENCE_WAIT           12
 #define DRM_VMW_OVERLAY              13
+#define DRM_VMW_CURSOR_BYPASS        14
 
 /*************************************************************************/
 /**
@@ -503,4 +504,35 @@ struct drm_vmw_overlay_arg {
 	struct drm_vmw_rect dst;
 };
 
+/*************************************************************************/
+/**
+ * DRM_VMW_CURSOR_BYPASS - Give extra information about cursor bypass.
+ *
+ */
+
+#define DRM_VMW_CURSOR_BYPASS_ALL    (1 << 0)
+#define DRM_VMW_CURSOR_BYPASS_FLAGS       (1)
+
+/**
+ * struct drm_vmw_cursor_bypass_arg
+ *
+ * @flags: Flags.
+ * @crtc_id: Crtc id, only used if DMR_CURSOR_BYPASS_ALL isn't passed.
+ * @xpos: X position of cursor.
+ * @ypos: Y position of cursor.
+ * @xhot: X hotspot.
+ * @yhot: Y hotspot.
+ *
+ * Argument to the DRM_VMW_CURSOR_BYPASS Ioctl.
+ */
+
+struct drm_vmw_cursor_bypass_arg {
+	uint32_t flags;
+	uint32_t crtc_id;
+	int32_t xpos;
+	int32_t ypos;
+	int32_t xhot;
+	int32_t yhot;
+};
+
 #endif
diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h b/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h
index 04d446a2dfb..db6b89b8bcd 100644
--- a/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h
+++ b/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h
@@ -44,6 +44,8 @@ struct vmw_driver
 {
     int fd;
 
+    void *cursor_priv;
+
     /* vmw_video.c */
     void *video_priv;
 };
@@ -69,6 +71,8 @@ Bool vmw_video_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw);
  * vmw_ioctl.c
  */
 
+int vmw_ioctl_cursor_bypass(struct vmw_driver *vmw, int xhot, int yhot);
+
 struct vmw_dma_buffer * vmw_ioctl_buffer_create(struct vmw_driver *vmw,
 						uint32_t size,
 						unsigned *handle);
diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c b/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c
index 3cac5b7760c..ad6993840d2 100644
--- a/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c
+++ b/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c
@@ -54,6 +54,23 @@ struct vmw_dma_buffer
     uint32_t size;
 };
 
+int
+vmw_ioctl_cursor_bypass(struct vmw_driver *vmw, int xhot, int yhot)
+{
+    struct drm_vmw_cursor_bypass_arg arg;
+    int ret;
+
+    memset(&arg, 0, sizeof(arg));
+    arg.flags = DRM_VMW_CURSOR_BYPASS_ALL;
+    arg.xhot = xhot;
+    arg.yhot = yhot;
+
+    ret = drmCommandWrite(vmw->fd, DRM_VMW_CURSOR_BYPASS,
+			  &arg, sizeof(arg));
+
+    return ret;
+}
+
 struct vmw_dma_buffer *
 vmw_ioctl_buffer_create(struct vmw_driver *vmw, uint32_t size, unsigned *handle)
 {
diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c b/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c
index 344ef0b3159..421906da996 100644
--- a/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c
+++ b/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c
@@ -33,16 +33,58 @@
 #include "vmw_hook.h"
 #include "vmw_driver.h"
 
+/* modified version of crtc functions */
+xf86CrtcFuncsRec vmw_screen_crtc_funcs;
+
+static void
+vmw_screen_cursor_load_argb(xf86CrtcPtr crtc, CARD32 *image)
+{
+    struct vmw_driver *vmw = modesettingPTR(crtc->scrn)->winsys_priv;
+    xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(crtc->scrn);
+    xf86CrtcFuncsPtr funcs = vmw->cursor_priv;
+    CursorPtr c = config->cursor;
+
+    /* Run the ioctl before uploading the image */
+    vmw_ioctl_cursor_bypass(vmw, c->bits->xhot, c->bits->yhot);
+
+    funcs->load_cursor_argb(crtc, image);
+}
+
+static void
+vmw_screen_cursor_init(ScrnInfoPtr pScrn, struct vmw_driver *vmw)
+{
+    xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn);
+    int i;
+
+    /* XXX assume that all crtc's have the same function struct */
+
+    /* Save old struct need to call the old functions as well */
+    vmw->cursor_priv = (void*)(config->crtc[0]->funcs);
+    memcpy(&vmw_screen_crtc_funcs, vmw->cursor_priv, sizeof(xf86CrtcFuncsRec));
+    vmw_screen_crtc_funcs.load_cursor_argb = vmw_screen_cursor_load_argb;
+
+    for (i = 0; i < config->num_crtc; i++)
+	config->crtc[i]->funcs = &vmw_screen_crtc_funcs;
+}
+
+static void
+vmw_screen_cursor_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw)
+{
+    xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn);
+    int i;
+
+    vmw_ioctl_cursor_bypass(vmw, 0, 0);
+
+    for (i = 0; i < config->num_crtc; i++)
+	config->crtc[i]->funcs = vmw->cursor_priv;
+}
+
 static Bool
 vmw_screen_init(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
     struct vmw_driver *vmw;
 
-    /* if gallium is used then we don't need to do anything. */
-    if (ms->screen)
-	return TRUE;
-
     vmw = xnfcalloc(sizeof(*vmw), 1);
     if (!vmw)
 	return FALSE;
@@ -50,6 +92,12 @@ vmw_screen_init(ScrnInfoPtr pScrn)
     vmw->fd = ms->fd;
     ms->winsys_priv = vmw;
 
+    vmw_screen_cursor_init(pScrn, vmw);
+
+    /* if gallium is used then we don't need to do anything more. */
+    if (ms->screen)
+	return TRUE;
+
     vmw_video_init(pScrn, vmw);
 
     return TRUE;
@@ -64,6 +112,8 @@ vmw_screen_close(ScrnInfoPtr pScrn)
     if (!vmw)
 	return TRUE;
 
+    vmw_screen_cursor_close(pScrn, vmw);
+
     vmw_video_close(pScrn, vmw);
 
     ms->winsys_priv = NULL;
diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_video.c b/src/gallium/winsys/drm/vmware/xorg/vmw_video.c
index 6e34aa21f38..d62c3b7296f 100644
--- a/src/gallium/winsys/drm/vmware/xorg/vmw_video.c
+++ b/src/gallium/winsys/drm/vmware/xorg/vmw_video.c
@@ -342,6 +342,8 @@ vmw_video_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw)
     debug_printf("%s: enter\n", __func__);
 
     video = vmw->video_priv;
+    if (!video)
+	return TRUE;
 
     for (i = 0; i < VMWARE_VID_NUM_PORTS; ++i) {
         vmw_video_port_cleanup(pScrn, &video->port[i]);
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index b6e0d823ed2..5615040946f 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -126,7 +126,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    case GL_RGB16:
       rb->Format = MESA_FORMAT_ARGB8888;
       rb->DataType = GL_UNSIGNED_BYTE;
-      irb->texformat = MESA_FORMAT_XRGB8888;
+      irb->texformat = MESA_FORMAT_ARGB8888; /* XXX: Need xrgb8888 */
       cpp = 4;
       break;
    case GL_RGBA:
@@ -314,6 +314,10 @@ intel_create_renderbuffer(gl_format format)
       irb->Base.DataType = GL_UNSIGNED_BYTE;
       break;
    case MESA_FORMAT_XRGB8888:
+      /* XXX this is a hack since XRGB surfaces don't seem to work
+       * properly yet.  Reading the alpha channel returns 0 instead of 1.
+       */
+      format = MESA_FORMAT_ARGB8888;
       irb->Base._BaseFormat = GL_RGB;
       irb->Base.DataType = GL_UNSIGNED_BYTE;
       break;
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c
index 309c90fdd0c..cf64d170ed6 100644
--- a/src/mesa/drivers/dri/r600/r700_assembler.c
+++ b/src/mesa/drivers/dri/r600/r700_assembler.c
@@ -4626,22 +4626,6 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm)
 	    need_barrier = GL_TRUE;
     }
 
-    switch (pAsm->pILInst[pAsm->uiCurInst].Opcode)
-    {
-        case OPCODE_TEX:
-            break;
-        case OPCODE_TXB:
-            radeon_error("do not support TXB yet\n");
-            return GL_FALSE;
-            break;
-        case OPCODE_TXP:
-            break;
-        default:
-            radeon_error("Internal error: bad texture op (not TEX)\n");
-            return GL_FALSE;
-            break;
-    }
-
     if (pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXP)
     {
         GLuint tmp = gethelpr(pAsm);
@@ -4820,7 +4804,18 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm)
 
     }
 
-    pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
+    if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB)
+    {
+        pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L;
+    }
+    else
+    {
+        pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
+    }
+
+    pAsm->is_tex = GL_TRUE;
+    if ( GL_TRUE == need_barrier )
+
     pAsm->is_tex = GL_TRUE;
     if ( GL_TRUE == need_barrier )
     {
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
index 8126777bf48..d7027400146 100644
--- a/src/mesa/drivers/dri/r600/r700_chip.c
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -1306,9 +1306,9 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(poly, always, 10, r700SendPolyState);
 	ALLOC_STATE(cb, cb, 18, r700SendCBState);
 	ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState);
+	ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState);
 	ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState);
 	ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState);
-	ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState);
 	ALLOC_STATE(sx, always, 9, r700SendSXState);
 	ALLOC_STATE(vgt, always, 41, r700SendVGTState);
 	ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState);
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index c81e80e820e..51fa6189377 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -262,29 +262,6 @@ void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h)
 	}
 }
 
-void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask )
-{
-   radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-   GLuint i;
-   drm_radeon_stipple_t stipple;
-
-   /* Must flip pattern upside down.
-   */
-   for ( i = 0 ; i < 32 ; i++ ) {
-      stipple.mask[31 - i] = ((GLuint *) mask)[i];
-   }
-
-   /* TODO: push this into cmd mechanism
-   */
-   radeon_firevertices(radeon);
-   LOCK_HARDWARE( radeon );
-
-   drmCommandWrite( radeon->dri.fd, DRM_RADEON_STIPPLE,
-	 &stipple, sizeof(stipple) );
-   UNLOCK_HARDWARE( radeon );
-}
-
-
 /* ================================================================
  * SwapBuffers with client-side throttling
  */
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
index def0cc17a91..0608fe2418c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -10,7 +10,6 @@ void radeonRecalcScissorRects(radeonContextPtr radeon);
 void radeonSetCliprects(radeonContextPtr radeon);
 void radeonUpdateScissor( GLcontext *ctx );
 void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h);
-void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask );
 
 void radeonWaitForIdleLocked(radeonContextPtr radeon);
 extern uint32_t radeonGetAge(radeonContextPtr radeon);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h
index 4e2c52c835c..12ab33a009c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_context.h
@@ -331,8 +331,12 @@ struct r100_hw_state {
 	struct radeon_state_atom stp;
 };
 
+struct radeon_stipple_state {
+	GLuint mask[32];
+};
 
 struct r100_state {
+	struct radeon_stipple_state stipple;
 	struct radeon_texture_state texture;
 };
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c
index 7ec641ff18a..fc21069a92c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_fbo.c
+++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c
@@ -369,6 +369,12 @@ radeon_framebuffer_renderbuffer(GLcontext * ctx,
 }
 
 
+/* TODO: According to EXT_fbo spec internal format of texture image
+ * once set during glTexImage call, should be preserved when
+ * attaching image to renderbuffer. When HW doesn't support
+ * rendering to format of attached image, set framebuffer
+ * completeness accordingly in radeon_validate_framebuffer (issue #79).
+ */
 static GLboolean
 radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, 
 		     struct gl_texture_image *texImage)
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
index bdbb9460bf0..a7f347202a1 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -437,9 +437,12 @@ static void migrate_image_to_miptree(radeon_mipmap_tree *mt,
 
 		radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel];
 
+		/* TODO: bring back these assertions once the FBOs are fixed */
+#if 0
 		assert(image->mtlevel == level);
 		assert(srclvl->size == dstlvl->size);
 		assert(srclvl->rowstride == dstlvl->rowstride);
+#endif
 
 		radeon_bo_map(image->mt->bo, GL_FALSE);
 
@@ -450,23 +453,18 @@ static void migrate_image_to_miptree(radeon_mipmap_tree *mt,
 
 		radeon_miptree_unreference(&image->mt);
 	} else {
-		/* need to confirm this value is correct */
-		if (_mesa_is_format_compressed(image->base.TexFormat)) {
-			unsigned size = _mesa_format_image_size(image->base.TexFormat,
-													image->base.Width,
-													image->base.Height,
-													image->base.Depth);
-			memcpy(dest, image->base.Data, size);
-		} else {
-			uint32_t srcrowstride;
-			uint32_t height;
+		const uint32_t srcrowstride = _mesa_format_row_stride(image->base.TexFormat, image->base.Width);
+		uint32_t rows = image->base.Height * image->base.Depth;
 
-			height = image->base.Height * image->base.Depth;
-			srcrowstride = image->base.Width * _mesa_get_format_bytes(image->base.TexFormat);
-			copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride,
-					height, srcrowstride);
+		if (_mesa_is_format_compressed(image->base.TexFormat)) {
+			uint32_t blockWidth, blockHeight;
+			_mesa_get_format_block_size(image->base.TexFormat, &blockWidth, &blockHeight);
+			rows = (rows + blockHeight - 1) / blockHeight;
 		}
 
+		copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride,
+				  rows, srcrowstride);
+
 		_mesa_free_texmemory(image->base.Data);
 		image->base.Data = 0;
 	}
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 4d0d35ee0cd..f6c733ab209 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -550,6 +550,31 @@ static void radeonPolygonOffset( GLcontext *ctx,
    rmesa->hw.zbs.cmd[ZBS_SE_ZBIAS_CONSTANT] = constant.ui32;
 }
 
+static void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask )
+{
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint i;
+   drm_radeon_stipple_t stipple;
+
+   /* Must flip pattern upside down.
+    */
+   for ( i = 0 ; i < 32 ; i++ ) {
+      rmesa->state.stipple.mask[31 - i] = ((GLuint *) mask)[i];
+   }
+
+   /* TODO: push this into cmd mechanism
+    */
+   radeon_firevertices(&rmesa->radeon);
+   LOCK_HARDWARE( &rmesa->radeon );
+
+   /* FIXME: Use window x,y offsets into stipple RAM.
+    */
+   stipple.mask = rmesa->state.stipple.mask;
+   drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_STIPPLE,
+		    &stipple, sizeof(drm_radeon_stipple_t) );
+   UNLOCK_HARDWARE( &rmesa->radeon );
+}
+
 static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
 {
    r100ContextPtr rmesa = R100_CONTEXT(ctx);
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c
index 0390d376ba2..28690325d12 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texture.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.c
@@ -515,15 +515,18 @@ static int image_matches_texture_obj(struct gl_texture_object *texObj,
 	struct gl_texture_image *texImage,
 	unsigned level)
 {
-	const struct gl_texture_image *baseImage = texObj->Image[0][level];
+	const struct gl_texture_image *baseImage = texObj->Image[0][texObj->BaseLevel];
+
+	if (!baseImage)
+		return 0;
 
 	if (level < texObj->BaseLevel || level > texObj->MaxLevel)
 		return 0;
 
 	const unsigned levelDiff = level - texObj->BaseLevel;
-	const unsigned refWidth = baseImage->Width >> levelDiff;
-	const unsigned refHeight = baseImage->Height >> levelDiff;
-	const unsigned refDepth = baseImage->Depth >> levelDiff;
+	const unsigned refWidth = MAX2(baseImage->Width >> levelDiff, 1);
+	const unsigned refHeight = MAX2(baseImage->Height >> levelDiff, 1);
+	const unsigned refDepth = MAX2(baseImage->Depth >> levelDiff, 1);
 
 	return (texImage->Width == refWidth &&
 			texImage->Height == refHeight &&