Index: Mesa/src/mesa/drivers/dri/radeon/radeon_context.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_context.c,v
retrieving revision 1.36
diff -u -r1.36 radeon_context.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_context.c	31 May 2005 04:04:24 -0000	1.36
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_context.c	5 Jun 2005 18:18:49 -0000
@@ -63,7 +63,7 @@
 #include "radeon_vtxfmt.h"
 #include "radeon_maos.h"
 
-#define DRIVER_DATE	"20050528"
+#define DRIVER_DATE	"20050605"
 
 #include "vblank.h"
 #include "utils.h"
@@ -338,7 +338,7 @@
 				 4,
 				 11, /* max 2D texture size is 2048x2048 */
 				 0,  /* 3D textures unsupported. */
-				 0,  /* cube textures unsupported. */
+				 9,  /* \todo: max cube texture size seems to be 512x512(x6) */
 				 11, /* max rect texture size is 2048x2048. */
 				 12,
 				 GL_FALSE );
@@ -412,6 +412,8 @@
    _math_matrix_set_identity( &rmesa->tmpmat );
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
+   if (rmesa->radeonScreen->drmSupportsCubeMaps)
+      _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
    if (rmesa->glCtx->Mesa_DXTn) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
       _mesa_enable_extension( ctx, "GL_S3_s3tc" );
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_context.h
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_context.h,v
retrieving revision 1.22
diff -u -r1.22 radeon_context.h
--- Mesa/src/mesa/drivers/dri/radeon/radeon_context.h	31 May 2005 04:04:24 -0000	1.22
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_context.h	5 Jun 2005 18:18:50 -0000
@@ -263,6 +263,16 @@
 #define TXR_PP_TEX_PITCH            2 /* 0x1d08, 0x1d10 for NPOT! */
 #define TXR_STATE_SIZE              3
 
+#define CUBE_CMD_0                  0
+#define CUBE_PP_CUBIC_FACES         1
+#define CUBE_CMD_1                  2
+#define CUBE_PP_CUBIC_OFFSET_0      3
+#define CUBE_PP_CUBIC_OFFSET_1      4
+#define CUBE_PP_CUBIC_OFFSET_2      5
+#define CUBE_PP_CUBIC_OFFSET_3      6
+#define CUBE_PP_CUBIC_OFFSET_4      7
+#define CUBE_STATE_SIZE             8
+
 #define ZBS_CMD_0              0
 #define ZBS_SE_ZBIAS_FACTOR             1
 #define ZBS_SE_ZBIAS_CONSTANT           2
@@ -414,6 +424,7 @@
    struct radeon_state_atom tcl;
    struct radeon_state_atom msc;
    struct radeon_state_atom tex[2];
+   struct radeon_state_atom cube[2];
    struct radeon_state_atom zbs;
    struct radeon_state_atom mtl; 
    struct radeon_state_atom mat[5]; 
@@ -534,6 +545,8 @@
    GLuint RenderIndex;
    GLuint vertex_size;
    GLuint vertex_format;
+   GLuint projformat;	/* bit n (0..2): Qn is used for projtex */
+
 
    struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
    GLuint vertex_attr_count;
@@ -605,10 +618,18 @@
    struct dynfn SecondaryColor3fvEXT;
    struct dynfn Normal3f;
    struct dynfn Normal3fv;
+#if 0		/* \todo: vtxfmt doesnt work correctly with cubemaps yet */
+   struct dynfn TexCoord3f;
+   struct dynfn TexCoord3fv;
+#endif
    struct dynfn TexCoord2f;
    struct dynfn TexCoord2fv;
    struct dynfn TexCoord1f;
    struct dynfn TexCoord1fv;
+#if 0
+   struct dynfn MultiTexCoord3fARB;
+   struct dynfn MultiTexCoord3fvARB;
+#endif
    struct dynfn MultiTexCoord2fARB;
    struct dynfn MultiTexCoord2fvARB;
    struct dynfn MultiTexCoord1fARB;
@@ -634,10 +655,18 @@
    struct dynfn *(*SecondaryColor3fvEXT)( GLcontext *, int );
    struct dynfn *(*Normal3f)( GLcontext *, int );
    struct dynfn *(*Normal3fv)( GLcontext *, int );
+#if 0
+   struct dynfn *(*TexCoord3f)( GLcontext *, int );
+   struct dynfn *(*TexCoord3fv)( GLcontext *, int );
+#endif
    struct dynfn *(*TexCoord2f)( GLcontext *, int );
    struct dynfn *(*TexCoord2fv)( GLcontext *, int );
    struct dynfn *(*TexCoord1f)( GLcontext *, int );
    struct dynfn *(*TexCoord1fv)( GLcontext *, int );
+#if 0
+   struct dynfn *(*MultiTexCoord3fARB)( GLcontext *, int );
+   struct dynfn *(*MultiTexCoord3fvARB)( GLcontext *, int );
+#endif
    struct dynfn *(*MultiTexCoord2fARB)( GLcontext *, int );
    struct dynfn *(*MultiTexCoord2fvARB)( GLcontext *, int );
    struct dynfn *(*MultiTexCoord1fARB)( GLcontext *, int );
@@ -652,23 +681,25 @@
    GLuint prim;
 };
 
+#define RADEON_MAX_VERTEX_SIZE 17
+
 struct radeon_vbinfo {
    GLint counter, initial_counter;
    GLint *dmaptr;
    void (*notify)( void );
    GLint vertex_size;
 
-   /* A maximum total of 15 elements per vertex:  3 floats for position, 3
+   /* A maximum total of 17 elements per vertex:  3 floats for position, 3
     * floats for normal, 4 floats for color, 4 bytes for secondary color,
-    * 2 floats for each texture unit (4 floats total).
+    * 3 floats for each texture unit (6 floats total).
     * 
-    * As soon as the 3rd TMU is supported or cube maps (or 3D textures) are
-    * supported, this value will grow.
+    * As soon as the 3rd TMU is supported this value will grow.
+    * If FPCOLOR/FPALPHA isnt used any more this value could shrink.
     * 
     * The position data is never actually stored here, so 3 elements could be
     * trimmed out of the buffer.
     */
-   union { float f; int i; radeon_color_t color; } vertex[15];
+   union { float f; int i; radeon_color_t color; } vertex[RADEON_MAX_VERTEX_SIZE];
 
    GLfloat *normalptr;
    GLfloat *floatcolorptr;
@@ -685,6 +716,7 @@
    GLboolean recheck;
    GLint nrverts;
    GLuint vertex_format;
+   GLuint projformat;	/* bit n (0..2): Qn is used for projtex */
 
    GLuint installed_vertex_format;
    GLuint installed_color_3f_sz;
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_ioctl.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_ioctl.c,v
retrieving revision 1.24
diff -u -r1.24 radeon_ioctl.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_ioctl.c	4 May 2005 20:11:38 -0000	1.24
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_ioctl.c	5 Jun 2005 18:18:51 -0000
@@ -151,6 +151,7 @@
    for (i = 0; i < mtu; ++i) {
        insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tex[i]);
        insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.txr[i]);
+       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.cube[i]);
    }
    insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.zbs);
    insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mtl);
@@ -422,6 +423,12 @@
    rmesa->ioctl.vertex_size = component[0]->aos_size;
    rmesa->ioctl.vertex_offset = 
       (component[0]->aos_start + offset * component[0]->aos_stride * 4);
+   if (RADEON_DEBUG & DEBUG_VERTS) {
+      fprintf( stderr, "%s: RADEON_OLD_PACKETS\n", __FUNCTION__);
+      fprintf( stderr, "  rmesa->ioctl.vertex_size= %d, rmesa->ioctl.vertex_offset= %d\n",
+               rmesa->ioctl.vertex_size,
+               rmesa->ioctl.vertex_offset);
+   }
 #else
    drm_radeon_cmd_header_t *cmd;
    int sz = AOS_BUFSZ(nr);
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h,v
retrieving revision 1.3
diff -u -r1.3 radeon_maos_vbtmp.h
--- Mesa/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h	24 Nov 2003 15:21:16 -0000	1.3
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h	5 Jun 2005 18:18:52 -0000
@@ -47,6 +47,7 @@
    GLuint tc0_stride, tc1_stride, col_stride, spec_stride, fog_stride;
    GLuint tc2_stride, norm_stride;
    GLuint fill_tex = 0;
+   GLuint cubemap_tex = 0;
    GLuint (*coord)[4];
    GLuint coord_stride; /* object coordinates */
    GLubyte dummy[4];
@@ -65,8 +66,11 @@
 	 const GLuint t2 = GET_TEXSOURCE(2);
 	 tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data;
 	 tc2_stride = VB->TexCoordPtr[t2]->stride;
-	 if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) {
-	    fill_tex |= (1<<2);
+	 if (DO_PTEX) {
+	    if (VB->TexCoordPtr[t2]->size < 3)
+	       fill_tex |= (1<<2);
+	    else if (VB->TexCoordPtr[t2]->size < 4)
+	       cubemap_tex |= (1<<2);
 	 }
       } else {
 	 tc2 = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_TEX2];
@@ -79,8 +83,11 @@
 	 const GLuint t1 = GET_TEXSOURCE(1);
 	 tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data;
 	 tc1_stride = VB->TexCoordPtr[t1]->stride;
-	 if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) {
-	    fill_tex |= (1<<1);
+	 if (DO_PTEX) {
+	    if (VB->TexCoordPtr[t1]->size < 3)
+	       fill_tex |= (1<<1);
+	    else if (VB->TexCoordPtr[t1]->size < 4)
+	       cubemap_tex |= (1<<1);
 	 }
       } else {
 	 tc1 = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_TEX1];
@@ -93,8 +100,11 @@
 	 const GLuint t0 = GET_TEXSOURCE(0);
 	 tc0_stride = VB->TexCoordPtr[t0]->stride;
 	 tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data;
-	 if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) {
-	    fill_tex |= (1<<0);
+	 if (DO_PTEX) {
+	    if (VB->TexCoordPtr[t0]->size < 3)
+	       fill_tex |= (1<<0);
+	    else if (VB->TexCoordPtr[t0]->size < 4)
+	       cubemap_tex |= (1<<0);
 	 }
       } else {
 	 tc0 = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_TEX0];
@@ -213,6 +223,8 @@
 	    if (DO_PTEX) {
 	       if (fill_tex & (1<<0))
 		  v[2].f = 1.0;
+	       else if (cubemap_tex & (1<<0))
+		  v[2].ui = tc0[0][2];	/* prefer cubemapping/3dtexures */
 	       else
 		  v[2].ui = tc0[0][3];
 	       if (TCL_DEBUG) fprintf(stderr, "%.2f ", v[2].f);
@@ -229,6 +241,8 @@
 	    if (DO_PTEX) {
 	       if (fill_tex & (1<<1))
 		  v[2].f = 1.0;
+	       else if (cubemap_tex & (1<<1))
+		  v[2].ui = tc1[0][2];
 	       else
 		  v[2].ui = tc1[0][3];
 	       if (TCL_DEBUG) fprintf(stderr, "%.2f ", v[2].f);
@@ -244,6 +258,8 @@
 	    if (DO_PTEX) {
 	       if (fill_tex & (1<<2))
 		  v[2].f = 1.0;
+	       else if (cubemap_tex & (1<<2))
+		  v[2].ui = tc2[0][2];
 	       else
 		  v[2].ui = tc2[0][3];
 	       v += 3;
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_maos_verts.c,v
retrieving revision 1.9
diff -u -r1.9 radeon_maos_verts.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_maos_verts.c	8 Apr 2004 08:54:24 -0000	1.9
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_maos_verts.c	5 Jun 2005 18:18:52 -0000
@@ -266,16 +266,21 @@
    if (inputs & VERT_BIT_TEX0) {
       req |= RADEON_CP_VC_FRMT_ST0;
 
-      if (VB->TexCoordPtr[0]->size == 4) {
+      if (RADEON_DEBUG & DEBUG_VERTS) 
+	 fprintf( stderr, "%s: st0 used\n", __FUNCTION__);
+
+      if (VB->TexCoordPtr[0]->size >= 3) {
 	 req |= RADEON_CP_VC_FRMT_Q0;
 	 vtx |= RADEON_TCL_VTX_Q0;
+	 if (RADEON_DEBUG & DEBUG_VERTS) 
+	    fprintf( stderr, "%s: q0 used\n", __FUNCTION__);
       }
    }
 
    if (inputs & VERT_BIT_TEX1) {
       req |= RADEON_CP_VC_FRMT_ST1;
 
-      if (VB->TexCoordPtr[1]->size == 4) {
+      if (VB->TexCoordPtr[1]->size >= 3) {
 	 req |= RADEON_CP_VC_FRMT_Q1;
 	 vtx |= RADEON_TCL_VTX_Q1;
       }
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_screen.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_screen.c,v
retrieving revision 1.32
diff -u -r1.32 radeon_screen.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_screen.c	4 May 2005 20:11:38 -0000	1.32
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_screen.c	5 Jun 2005 18:18:53 -0000
@@ -255,6 +255,7 @@
 	    fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_IRQ_NR): %d\n", ret);
 	    return NULL;
 	 }
+	 screen->drmSupportsCubeMaps = (sPriv->drmMinor >= 15);
       }
    }
 
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_screen.h
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_screen.h,v
retrieving revision 1.14
diff -u -r1.14 radeon_screen.h
--- Mesa/src/mesa/drivers/dri/radeon/radeon_screen.h	4 May 2005 20:11:38 -0000	1.14
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_screen.h	5 Jun 2005 18:18:53 -0000
@@ -98,6 +98,7 @@
 
    /* Configuration cache with default values for all contexts */
    driOptionCache optionCache;
+   GLboolean drmSupportsCubeMaps;
 } radeonScreenRec, *radeonScreenPtr;
 
 #endif /* __RADEON_SCREEN_H__ */
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_state_init.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_state_init.c,v
retrieving revision 1.14
diff -u -r1.14 radeon_state_init.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_state_init.c	16 Feb 2005 15:02:25 -0000	1.14
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_state_init.c	5 Jun 2005 18:18:54 -0000
@@ -113,8 +113,11 @@
 
 
 CHECK( always, GL_TRUE )
+CHECK( never, GL_FALSE )
 CHECK( tex0, ctx->Texture.Unit[0]._ReallyEnabled )
 CHECK( tex1, ctx->Texture.Unit[1]._ReallyEnabled )
+CHECK( cube0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT))
+CHECK( cube1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT))
 CHECK( fog, ctx->Fog.Enabled )
 TCL_CHECK( tcl, GL_TRUE )
 TCL_CHECK( tcl_tex0, ctx->Texture.Unit[0]._ReallyEnabled )
@@ -230,6 +233,16 @@
    ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 );
    ALLOC_STATE( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0 );
    ALLOC_STATE( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0 );
+   if (rmesa->radeonScreen->drmSupportsCubeMaps)
+   {
+      ALLOC_STATE( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
+      ALLOC_STATE( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
+   }
+   else
+   {
+      ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
+      ALLOC_STATE( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
+   }
    ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 );
    ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 );
    ALLOC_STATE( mat[2], tcl_eyespace_or_lighting, MAT_STATE_SIZE, "MAT/it-modelview", 1 );
@@ -270,6 +283,10 @@
    rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_0);
    rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_1);
    rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_1);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T1);
    rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
    rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT);
    rmesa->hw.mtl.cmd[MTL_CMD_0] = 
@@ -310,8 +327,8 @@
 				     RADEON_CHROMA_FUNC_FAIL |
 				     RADEON_CHROMA_KEY_NEAREST |
 				     RADEON_SHADOW_FUNC_EQUAL |
-				     RADEON_SHADOW_PASS_1 |
-				     RADEON_RIGHT_HAND_CUBE_OGL);
+				     RADEON_SHADOW_PASS_1);
+     /* isn't it strange that we mustn't set RADEON_RIGHT_HAND_CUBE_OGL ? */
 
    rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] = (RADEON_FOG_VERTEX |
 					  RADEON_FOG_USE_DEPTH);
@@ -478,6 +495,18 @@
 	   RADEON_SCALE_1X |
 	   RADEON_CLAMP_TX);
       rmesa->hw.tex[i].cmd[TEX_PP_TFACTOR] = 0;
+
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_0] =
+	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_1] =
+	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_2] =
+	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_3] =
+	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_4] =
+	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
    }
 
    /* Can only add ST1 at the time of doing some multitex but can keep
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_swtcl.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_swtcl.c,v
retrieving revision 1.20
diff -u -r1.20 radeon_swtcl.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_swtcl.c	31 May 2005 04:13:45 -0000	1.20
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_swtcl.c	5 Jun 2005 18:18:55 -0000
@@ -94,6 +94,7 @@
    struct vertex_buffer *VB = &tnl->vb;
    GLuint index = tnl->render_inputs;
    int fmt_0 = 0;
+   int projtex = 0;
    int offset = 0;
 
 
@@ -184,13 +185,17 @@
 	    switch (sz) {
 	    case 1:
 	    case 2:
-	    case 3:
 	       EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_2F,
 			  radeon_cp_vc_frmts[i][0] );
 	       break;
-	    case 4:
+	    case 3:
+	       EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_3F,
+			  radeon_cp_vc_frmts[i][1] );
+	       break;
+	    case 4:	/* FIXME: what should we do if its still cubemaps ? */
 	       EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_3F_XYW,
 			  radeon_cp_vc_frmts[i][1] );
+	       projtex |= 1<<i;
 	       break;
 	    default:
 	       continue;
@@ -200,9 +205,11 @@
    }
 
    if ( rmesa->tnl_index != index ||
-	fmt_0 != rmesa->swtcl.vertex_format) {
+	fmt_0 != rmesa->swtcl.vertex_format ||
+	projtex != rmesa->swtcl.projformat) {
       RADEON_NEWPRIM(rmesa);
       rmesa->swtcl.vertex_format = fmt_0;
+      rmesa->swtcl.projformat= projtex;
       rmesa->swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
 			      rmesa->swtcl.vertex_attrs, 
@@ -607,7 +614,8 @@
 
 #define RADEON_TWOSIDE_BIT	0x01
 #define RADEON_UNFILLED_BIT	0x02
-#define RADEON_MAX_TRIFUNC	0x08
+/* FIXME: why 8 but not 4 (as in r200 driver) ? */
+#define RADEON_MAX_TRIFUNC	0x04
 
 
 static struct {
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.c,v
retrieving revision 1.13
diff -u -r1.13 radeon_tcl.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.c	16 May 2005 08:36:28 -0000	1.13
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.c	5 Jun 2005 18:18:55 -0000
@@ -240,6 +240,9 @@
 			  GLuint last,
 			  GLuint flags )
 {
+   if( RADEON_DEBUG & DEBUG_PRIMS)
+      fprintf( stderr, "%s: flags= %0x, prim:%s\n",
+       __FUNCTION__, flags, _mesa_lookup_enum_by_nr( flags & PRIM_MODE_MASK ));
    tcl_render_tab_verts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
 }
 
@@ -248,6 +251,8 @@
 			     GLuint last,
 			     GLuint flags )
 {
+   if( RADEON_DEBUG & DEBUG_PRIMS)
+      fprintf( stderr, "%s: flags= %0x\n", __FUNCTION__, flags);
    tcl_render_tab_elts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
 }
 
@@ -444,7 +449,10 @@
    "Texgen unit 0",
    "Texgen unit 1",
    "Texgen unit 2",
-   "User disable"
+   "User disable",
+   "texture rectangle unit 0",
+   "texture rectangle unit 1",
+   "texture rectangle unit 2"
 };
 
 
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.h
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.h,v
retrieving revision 1.4
diff -u -r1.4 radeon_tcl.h
--- Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.h	31 Jan 2005 23:40:06 -0000	1.4
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_tcl.h	5 Jun 2005 18:18:55 -0000
@@ -59,7 +59,7 @@
 #define RADEON_TCL_FALLBACK_TEXRECT_1         0x200 /* texture rectangle */
 #define RADEON_TCL_FALLBACK_TEXRECT_2         0x400 /* texture rectangle */
 
-#define RADEON_MAX_TCL_VERTSIZE (15*4)
+#define RADEON_MAX_TCL_VERTSIZE (RADEON_MAX_VERTEX_SIZE*4)
 
 #define TCL_FALLBACK( ctx, bit, mode )	radeonTclFallback( ctx, bit, mode )
 
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_tex.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_tex.c,v
retrieving revision 1.15
diff -u -r1.15 radeon_tex.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_tex.c	7 Oct 2004 23:30:30 -0000	1.15
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_tex.c	5 Jun 2005 18:18:56 -0000
@@ -161,6 +161,31 @@
 }
 
 /**
+ * returns a non-mipmap texfilter out of any texfilter:
+ * ugly workaround since mipmaped cubemaps seem to be unsupported on radeon.
+ * I think the mipmaps are generated but are unused.. which is worse when
+ * using dynamically generated textures (reflections, etc)...
+ *
+ * \todo: fix mipmaped cubemaps and get rid of this.
+ *
+ * \param filtermode original filtermode
+ */
+static GLenum radeonFakedCubeTexFilter( GLenum filtermode)
+{
+   switch (filtermode) {
+      case GL_NEAREST_MIPMAP_NEAREST:
+      case GL_NEAREST_MIPMAP_LINEAR:
+	 return (GL_NEAREST);
+      case GL_LINEAR_MIPMAP_NEAREST:
+      case GL_LINEAR_MIPMAP_LINEAR:
+	 return (GL_LINEAR);
+      default:
+	 break;
+   }
+   return (filtermode);
+}
+
+/**
  * Set the texture magnification and minification modes.
  * 
  * \param t Texture whose filter modes are to be set
@@ -260,7 +285,10 @@
 
       radeonSetTexWrap( t, texObj->WrapS, texObj->WrapT );
       radeonSetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+      if (texObj->Target != GL_TEXTURE_CUBE_MAP)
+	 radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+      else
+	 radeonSetTexFilter( t, radeonFakedCubeTexFilter(texObj->MinFilter), texObj->MagFilter );
       radeonSetTexBorderColor( t, texObj->_BorderChan );
    }
 
@@ -731,7 +759,10 @@
    case GL_TEXTURE_MAG_FILTER:
    case GL_TEXTURE_MAX_ANISOTROPY_EXT:
       radeonSetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+      if (target != GL_TEXTURE_CUBE_MAP)
+	 radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+      else
+	 radeonSetTexFilter( t, radeonFakedCubeTexFilter( texObj->MinFilter), texObj->MagFilter );
       break;
 
    case GL_TEXTURE_WRAP_S:
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_texmem.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_texmem.c,v
retrieving revision 1.11
diff -u -r1.11 radeon_texmem.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_texmem.c	10 Feb 2005 22:36:07 -0000	1.11
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_texmem.c	5 Jun 2005 18:18:56 -0000
@@ -68,6 +68,7 @@
 	 if ( t == rmesa->state.texture.unit[i].texobj ) {
 	    rmesa->state.texture.unit[i].texobj = NULL;
 	    rmesa->hw.tex[i].dirty = GL_FALSE;
+	    rmesa->hw.cube[i].dirty = GL_FALSE;
 	 }
       }
    }
@@ -226,7 +227,7 @@
    imageWidth = texImage->Width;
    imageHeight = texImage->Height;
 
-   offset = t->bufAddr;
+   offset = t->bufAddr + t->base.totalSize * face / 6;
 
    if ( RADEON_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
       GLint imageX = 0;
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_texstate.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_texstate.c,v
retrieving revision 1.15
diff -u -r1.15 radeon_texstate.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_texstate.c	28 May 2005 23:45:39 -0000	1.15
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_texstate.c	5 Jun 2005 18:19:00 -0000
@@ -284,6 +284,22 @@
     */
    t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
 
+   /* Setup remaining cube face blits, if needed */
+   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
+      const GLuint faceSize = t->base.totalSize;
+      GLuint face;
+      /* reuse face 0 x/y/width/height - just update the offset when uploading */
+      for (face = 1; face < 6; face++) {
+         for (i = 0; i < numLevels; i++) {
+            t->image[face][i].x =  t->image[0][i].x;
+            t->image[face][i].y =  t->image[0][i].y;
+            t->image[face][i].width  = t->image[0][i].width;
+            t->image[face][i].height = t->image[0][i].height;
+         }
+      }
+      t->base.totalSize = 6 * faceSize; /* total texmem needed */
+   }
+
    /* Hardware state:
     */
    t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK;
@@ -291,10 +307,27 @@
 
    t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK |
 		       RADEON_TXFORMAT_HEIGHT_MASK |
-                       RADEON_TXFORMAT_CUBIC_MAP_ENABLE);
+                       RADEON_TXFORMAT_CUBIC_MAP_ENABLE |
+                       RADEON_TXFORMAT_F5_WIDTH_MASK |
+                       RADEON_TXFORMAT_F5_HEIGHT_MASK);
    t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) |
 		      (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT));
 
+   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
+      assert(log2Width == log2Height);
+      t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) |
+                         (log2Height << RADEON_TXFORMAT_F5_HEIGHT_SHIFT) |
+                         (RADEON_TXFORMAT_CUBIC_MAP_ENABLE));
+      t->pp_cubic_faces = ((log2Width << RADEON_FACE_WIDTH_1_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_1_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_2_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_2_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_3_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_3_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_4_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_4_SHIFT));
+   }
+
    t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) |
                    ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16));
 
@@ -816,30 +849,63 @@
    cmd[TEX_PP_TXFORMAT] |= texobj->pp_txformat & TEXOBJ_TXFORMAT_MASK;
    cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset;
    cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color;
-   RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.tex[unit] );
 
-   if (texobj->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
+   if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) {
+      GLuint *cube_cmd = RADEON_DB_STATE( cube[unit] );
+      GLuint bytesPerFace = texobj->base.totalSize / 6;
+      ASSERT(texobj->totalSize % 6 == 0);
+
+      cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
+      /* dont know if this setup conforms to OpenGL.. 
+       * at least it matches the behavior of mesa software renderer
+       */
+      cube_cmd[CUBE_PP_CUBIC_OFFSET_0] = texobj->pp_txoffset; /* right */
+      cube_cmd[CUBE_PP_CUBIC_OFFSET_1] = texobj->pp_txoffset + 1 * bytesPerFace; /* left */
+      cube_cmd[CUBE_PP_CUBIC_OFFSET_2] = texobj->pp_txoffset + 2 * bytesPerFace; /* top */
+      cube_cmd[CUBE_PP_CUBIC_OFFSET_3] = texobj->pp_txoffset + 3 * bytesPerFace; /* bottom */
+      cube_cmd[CUBE_PP_CUBIC_OFFSET_4] = texobj->pp_txoffset + 4 * bytesPerFace; /* front */
+      RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.cube[unit] );
+      cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset + 5 * bytesPerFace; /* back */
+   }
+   else if (texobj->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
       GLuint *txr_cmd = RADEON_DB_STATE( txr[unit] );
       txr_cmd[TXR_PP_TEX_SIZE] = texobj->pp_txsize; /* NPOT only! */
       txr_cmd[TXR_PP_TEX_PITCH] = texobj->pp_txpitch; /* NPOT only! */
       RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.txr[unit] );
    }
 
+   RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.tex[unit] );
+
    texobj->dirty_state &= ~(1<<unit);
 }
 
 
 
+/* FIXME: cubemap-patch: I didnt bother trying to understand the following, so just
+ * copied from R200 and/or earlier (nonworking) cubemap-patches I found.
+ * Maybe its only needed if cubemapping is done with hw-tcl.
+ *
+ * btw: is cubemapping even possible in hw-tcl on R100 ?
+ * Or is it only possible if in non-texgen-mode?
+ * Or do we need texgen-mode for Q first? (its a fallback atm)
+ * or some magic between texgen-Q and R ?
+ */
 
 static void set_texgen_matrix( radeonContextPtr rmesa, 
 			       GLuint unit,
 			       const GLfloat *s_plane,
-			       const GLfloat *t_plane )
+			       const GLfloat *t_plane,
+			       const GLfloat *r_plane,
+			       const GLfloat *q_plane)
 {
    static const GLfloat scale_identity[4] = { 1,1,1,1 };
 
+/*
    if (!TEST_EQ_4V( s_plane, scale_identity) ||
-       !TEST_EQ_4V( t_plane, scale_identity)) {
+       !TEST_EQ_4V( t_plane, scale_identity) ||
+       !TEST_EQ_4V( r_plane, scale_identity) ||
+       !TEST_EQ_4V( q_plane, scale_identity))		*/	 {
+
       rmesa->TexGenEnabled |= RADEON_TEXMAT_0_ENABLE<<unit;
       rmesa->TexGenMatrix[unit].m[0]  = s_plane[0];
       rmesa->TexGenMatrix[unit].m[4]  = s_plane[1];
@@ -850,10 +916,60 @@
       rmesa->TexGenMatrix[unit].m[5]  = t_plane[1];
       rmesa->TexGenMatrix[unit].m[9]  = t_plane[2];
       rmesa->TexGenMatrix[unit].m[13] = t_plane[3];
+
+      /* NOTE: r_plane goes in the 4th row, not 3rd! */
+      rmesa->TexGenMatrix[unit].m[3]  = r_plane[0];
+      rmesa->TexGenMatrix[unit].m[7]  = r_plane[1];
+      rmesa->TexGenMatrix[unit].m[11] = r_plane[2];
+      rmesa->TexGenMatrix[unit].m[15] = r_plane[3];
+
+
+/* FIXME: check this: atm we overwrite the r_plane */
+
+      rmesa->TexGenMatrix[unit].m[2]  = 0;
+      rmesa->TexGenMatrix[unit].m[6]  = 0;
+      rmesa->TexGenMatrix[unit].m[10] = 0;
+      rmesa->TexGenMatrix[unit].m[14] = 0;
+
+
+      rmesa->TexGenMatrix[unit].m[3]  = 0;
+      rmesa->TexGenMatrix[unit].m[7]  = 0;
+      rmesa->TexGenMatrix[unit].m[11] = 0;
+      rmesa->TexGenMatrix[unit].m[15] = 0;
+
       rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 }
 
+/* Need this special matrix to get correct reflection map coords */
+static void
+set_texgen_reflection_matrix( radeonContextPtr rmesa, GLuint unit )
+{
+   static const GLfloat m[16] = {
+      -1,  0,  0,  0,
+       0, -1,  0,  0,
+       0,  0,  0, -1,
+       0,  0, -1,  0 };
+   _math_matrix_loadf( &(rmesa->TexGenMatrix[unit]), m);
+   _math_matrix_analyse( &(rmesa->TexGenMatrix[unit]) );
+   rmesa->TexGenEnabled |= RADEON_TEXMAT_0_ENABLE<<unit;
+}
+
+/* Need this special matrix to get correct normal map coords */
+static void
+set_texgen_normal_map_matrix( radeonContextPtr rmesa, GLuint unit )
+{
+   static const GLfloat m[16] = {
+      1, 0, 0, 0,
+      0, 1, 0, 0,
+      0, 0, 0, 1,
+      0, 0, 1, 0 };
+   _math_matrix_loadf( &(rmesa->TexGenMatrix[unit]), m);
+   _math_matrix_analyse( &(rmesa->TexGenMatrix[unit]) );
+   rmesa->TexGenEnabled |= RADEON_TEXMAT_0_ENABLE<<unit;
+}
+
+
 /* Ignoring the Q texcoord for now.
  *
  * Returns GL_FALSE if fallback required.  
@@ -870,13 +986,15 @@
    rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_INPUT_MASK<<inputshift);
    rmesa->TexGenNeedNormals[unit] = 0;
 
-   if ((texUnit->TexGenEnabled & (S_BIT|T_BIT)) == 0) {
+   if ((texUnit->TexGenEnabled & (S_BIT|T_BIT|R_BIT)) == 0) {
       /* Disabled, no fallback:
        */
+/* FIXME: shouldnt we disable texgen? */
       rmesa->TexGenEnabled |= 
 	 (RADEON_TEXGEN_INPUT_TEXCOORD_0+unit) << inputshift;
       return GL_TRUE;
    }
+#if 1
    else if (texUnit->TexGenEnabled & Q_BIT) {
       /* Very easy to do this, in fact would remove a fallback case
        * elsewhere, but I haven't done it yet...  Fallback: 
@@ -885,40 +1003,71 @@
 	fprintf(stderr, "fallback Q_BIT\n");
       return GL_FALSE;
    }
-   else if ((texUnit->TexGenEnabled & (S_BIT|T_BIT)) != (S_BIT|T_BIT) ||
-	    texUnit->GenModeS != texUnit->GenModeT) {
+#endif
+   else if (texUnit->TexGenEnabled == (S_BIT|T_BIT) &&
+	    texUnit->GenModeS == texUnit->GenModeT) {
+      /* OK */
+      rmesa->TexGenEnabled |= RADEON_TEXGEN_TEXMAT_0_ENABLE << unit;
+      /* continue */
+   }
+#if 0
+/* doesnt work at the moment */
+   else if (texUnit->TexGenEnabled == (S_BIT|T_BIT|R_BIT) &&
+	    texUnit->GenModeS == texUnit->GenModeT &&
+            texUnit->GenModeT == texUnit->GenModeR) {
+      /* OK */
+      rmesa->TexGenEnabled |= RADEON_TEXGEN_TEXMAT_0_ENABLE << unit;
+      /* continue */
+   }
+#endif
+/* FIXME: dont know if Q_BIT works ... */
+/* just ignore the R texgenmode */
+   else if (texUnit->TexGenEnabled == (S_BIT|T_BIT|R_BIT|Q_BIT) &&
+	    texUnit->GenModeS == texUnit->GenModeT &&
+            texUnit->GenModeT == texUnit->GenModeQ) {
+      /* OK */
+      rmesa->TexGenEnabled |= RADEON_TEXGEN_TEXMAT_0_ENABLE << unit;
+      /* continue */
+   }
+   else {
       /* Mixed modes, fallback:
        */
       if (RADEON_DEBUG & DEBUG_FALLBACKS) 
         fprintf(stderr, "fallback mixed texgen\n");
       return GL_FALSE;
    }
-   else
-      rmesa->TexGenEnabled |= RADEON_TEXGEN_TEXMAT_0_ENABLE << unit;
+
+   rmesa->TexGenEnabled |= RADEON_TEXGEN_TEXMAT_0_ENABLE << unit;
 
    switch (texUnit->GenModeS) {
    case GL_OBJECT_LINEAR:
       rmesa->TexGenEnabled |= RADEON_TEXGEN_INPUT_OBJ << inputshift;
       set_texgen_matrix( rmesa, unit, 
 			 texUnit->ObjectPlaneS,
-			 texUnit->ObjectPlaneT);
+			 texUnit->ObjectPlaneT,
+                         texUnit->ObjectPlaneR,
+                         texUnit->ObjectPlaneQ);
       break;
 
    case GL_EYE_LINEAR:
       rmesa->TexGenEnabled |= RADEON_TEXGEN_INPUT_EYE << inputshift;
       set_texgen_matrix( rmesa, unit, 
 			 texUnit->EyePlaneS,
-			 texUnit->EyePlaneT);
+			 texUnit->EyePlaneT,
+			 texUnit->EyePlaneR,
+			 texUnit->EyePlaneQ);
       break;
 
    case GL_REFLECTION_MAP_NV:
       rmesa->TexGenNeedNormals[unit] = GL_TRUE;
       rmesa->TexGenEnabled |= RADEON_TEXGEN_INPUT_EYE_REFLECT<<inputshift;
+      set_texgen_reflection_matrix(rmesa, unit);
       break;
 
    case GL_NORMAL_MAP_NV:
       rmesa->TexGenNeedNormals[unit] = GL_TRUE;
       rmesa->TexGenEnabled |= RADEON_TEXGEN_INPUT_EYE_NORMAL<<inputshift;
+      set_texgen_normal_map_matrix(rmesa, unit);
       break;
 
    case GL_SPHERE_MAP:
@@ -1024,6 +1173,48 @@
    return GL_TRUE;
 }
 
+static GLboolean enable_tex_cube( GLcontext *ctx, int unit )
+{
+   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *tObj = texUnit->_Current;
+   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
+   GLuint face;
+
+   /* Need to load the 2d images associated with this unit.
+    */
+   if (t->pp_txformat & RADEON_TXFORMAT_NON_POWER2) {
+      t->pp_txformat &= ~RADEON_TXFORMAT_NON_POWER2;
+      for (face = 0; face < 6; face++)
+         t->base.dirty_images[face] = ~0;
+   }
+
+   ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
+
+   if ( t->base.dirty_images[0] || t->base.dirty_images[1] ||
+        t->base.dirty_images[2] || t->base.dirty_images[3] ||
+        t->base.dirty_images[4] || t->base.dirty_images[5] ) {
+      /* flush */
+      RADEON_FIREVERTICES( rmesa );
+      /* layout memory space, once for all faces */
+      radeonSetTexImages( rmesa, tObj );
+   }
+
+   /* upload (per face) */
+   for (face = 0; face < 6; face++) {
+      if (t->base.dirty_images[face]) {
+         radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, face );
+      }
+   }
+      
+   if ( !t->base.memBlock ) {
+      /* texmem alloc failed, use s/w fallback */
+      return GL_FALSE;
+   }
+
+   return GL_TRUE;
+}
+
 static GLboolean enable_tex_rect( GLcontext *ctx, int unit )
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
@@ -1146,6 +1337,10 @@
       return (enable_tex_2d( ctx, unit ) &&
 	      update_tex_common( ctx, unit ));
    }
+   else if ( texUnit->_ReallyEnabled & (TEXTURE_CUBE_BIT) ) {
+      return (enable_tex_cube( ctx, unit ) &&
+	      update_tex_common( ctx, unit ));
+   }
    else if ( texUnit->_ReallyEnabled ) {
       return GL_FALSE;
    }
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c,v
retrieving revision 1.9
diff -u -r1.9 radeon_vtxfmt.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c	16 Feb 2005 15:02:25 -0000	1.9
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c	5 Jun 2005 18:19:01 -0000
@@ -87,10 +87,18 @@
    count_func( "SecondaryColor3ubv", &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
    count_func( "Normal3f", &rmesa->vb.dfn_cache.Normal3f );
    count_func( "Normal3fv", &rmesa->vb.dfn_cache.Normal3fv );
+#if 0		/* \todo: make this working */
+   count_func( "TexCoord3f", &rmesa->vb.dfn_cache.TexCoord3f );
+   count_func( "TexCoord3fv", &rmesa->vb.dfn_cache.TexCoord3fv );
+#endif
    count_func( "TexCoord2f", &rmesa->vb.dfn_cache.TexCoord2f );
    count_func( "TexCoord2fv", &rmesa->vb.dfn_cache.TexCoord2fv );
    count_func( "TexCoord1f", &rmesa->vb.dfn_cache.TexCoord1f );
    count_func( "TexCoord1fv", &rmesa->vb.dfn_cache.TexCoord1fv );
+#if 0
+   count_func( "MultiTexCoord3fARB", &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
+   count_func( "MultiTexCoord3fvARB", &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
+#endif
    count_func( "MultiTexCoord2fARB", &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
    count_func( "MultiTexCoord2fvARB", &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
    count_func( "MultiTexCoord1fARB", &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
@@ -135,14 +143,22 @@
    if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_ST0) {
       ctx->Current.Attrib[VERT_ATTRIB_TEX0][0] = rmesa->vb.texcoordptr[0][0];
       ctx->Current.Attrib[VERT_ATTRIB_TEX0][1] = rmesa->vb.texcoordptr[0][1];
-      ctx->Current.Attrib[VERT_ATTRIB_TEX0][2] = 0.0F;
+      if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_Q0) {
+	 ctx->Current.Attrib[VERT_ATTRIB_TEX0][2] = rmesa->vb.texcoordptr[0][2];
+      } else {
+      	 ctx->Current.Attrib[VERT_ATTRIB_TEX0][2] = 0.0F;
+      }
       ctx->Current.Attrib[VERT_ATTRIB_TEX0][3] = 1.0F;
    }
 
    if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_ST1) {
       ctx->Current.Attrib[VERT_ATTRIB_TEX1][0] = rmesa->vb.texcoordptr[1][0];
       ctx->Current.Attrib[VERT_ATTRIB_TEX1][1] = rmesa->vb.texcoordptr[1][1];
-      ctx->Current.Attrib[VERT_ATTRIB_TEX1][2] = 0.0F;
+      if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_Q1) {
+	 ctx->Current.Attrib[VERT_ATTRIB_TEX1][2] = rmesa->vb.texcoordptr[1][2];
+      } else {
+      	 ctx->Current.Attrib[VERT_ATTRIB_TEX1][2] = 0.0F;
+      }
       ctx->Current.Attrib[VERT_ATTRIB_TEX1][3] = 1.0F;
    }
 
@@ -180,6 +196,13 @@
    rmesa->tcl.nr_aos_components = 1;
    rmesa->dma.flush = NULL;
 
+   if( RADEON_DEBUG & DEBUG_PRIMS) {
+      fprintf( stderr, "%s: \n", __FUNCTION__);
+      fprintf( stderr, "  rmesa->tcl.vertex_format=0x%0x\n", rmesa->tcl.vertex_format);
+      fprintf( stderr, "  rmesa->vb.vertex_size=%d\n", rmesa->vb.vertex_size);
+      fprintf( stderr, "  rmesa->vb.nrprims=%d\n", rmesa->vb.nrprims);
+   }
+      
    /* Optimize the primitive list:
     */
    if (rmesa->vb.nrprims > 1) {
@@ -199,6 +222,10 @@
       rmesa->vb.nrprims = j+1;
    }
 
+   if( RADEON_DEBUG & DEBUG_PRIMS) {
+      fprintf( stderr, "  rmesa->vb.nrprims=%d\n", rmesa->vb.nrprims);
+   }
+
    for (i = 0 ; i < rmesa->vb.nrprims; i++) {
       if (RADEON_DEBUG & DEBUG_PRIMS)
 	 fprintf(stderr, "vtxfmt prim %d: %s %d..%d\n", i,
@@ -215,6 +242,9 @@
 
    rmesa->vb.nrprims = 0;
    radeonReleaseDmaRegion( rmesa, &tmp, __FUNCTION__ );
+
+   if( RADEON_DEBUG & DEBUG_PRIMS)
+      fprintf( stderr, "%s: --end--\n", __FUNCTION__);
 }
 
 
@@ -262,7 +292,7 @@
  * memory.  Could also use the counter/notify mechanism to populate
  * tmp on the fly as vertices are generated.  
  */
-static GLuint copy_dma_verts( radeonContextPtr rmesa, GLfloat (*tmp)[15] )
+static GLuint copy_dma_verts( radeonContextPtr rmesa, GLfloat (*tmp)[RADEON_MAX_VERTEX_SIZE] )
 {
    GLuint ovf, i;
    GLuint nr = (rmesa->vb.initial_counter - rmesa->vb.counter) - rmesa->vb.primlist[rmesa->vb.nrprims].start;
@@ -350,11 +380,11 @@
 }
 
 
-static void VFMT_FALLBACK( const char *caller )
+void VFMT_FALLBACK( const char *caller )
 {
    GET_CURRENT_CONTEXT(ctx);
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][15];
+   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
    GLuint i, prim;
    GLuint ind = rmesa->vb.vertex_format;
    GLuint nrverts;
@@ -422,13 +452,26 @@
       }
 
       if (ind & RADEON_CP_VC_FRMT_ST0) {
-	 GL_CALL(TexCoord2fv)( &tmp[i][offset] ); 
-	 offset += 2;
+/* FIXME: what should happen if Q0 is set from projective texture? */
+/* unfortunately the vertex_format used here is from the vb struct, not from the swtcl struct */
+/* so I think we cant use the projformat from swtcl */
+      	 if (ind & RADEON_CP_VC_FRMT_Q0) {
+      	    GL_CALL(TexCoord3fv)( &tmp[i][offset] ); 
+      	    offset += 3;
+      	 } else {
+	    GL_CALL(TexCoord2fv)( &tmp[i][offset] ); 
+	    offset += 2;
+	 }
       }
 
       if (ind & RADEON_CP_VC_FRMT_ST1) {
-	 GL_CALL(MultiTexCoord2fvARB)( GL_TEXTURE1_ARB, &tmp[i][offset] );
-	 offset += 2;
+      	 if (ind & RADEON_CP_VC_FRMT_Q1) {
+	    GL_CALL(MultiTexCoord3fvARB)( GL_TEXTURE1, &tmp[i][offset] );
+	    offset += 3;
+	 } else {
+	    GL_CALL(MultiTexCoord2fvARB)( GL_TEXTURE1, &tmp[i][offset] );
+	    offset += 2;
+	 }
       }
       GL_CALL(Vertex3fv)( &tmp[i][0] );
    }
@@ -455,11 +498,21 @@
    if (ind & RADEON_CP_VC_FRMT_PKSPEC) 
        GL_CALL(SecondaryColor3ubEXT)( rmesa->vb.specptr->red, rmesa->vb.specptr->green, rmesa->vb.specptr->blue ); 
 
-   if (ind & RADEON_CP_VC_FRMT_ST0) 
-      GL_CALL(TexCoord2fv)( rmesa->vb.texcoordptr[0] );
+   if (ind & RADEON_CP_VC_FRMT_ST0) {
+      if (ind & RADEON_CP_VC_FRMT_Q0) {
+      	 GL_CALL(TexCoord3fv)( rmesa->vb.texcoordptr[0] );
+      } else {
+      	 GL_CALL(TexCoord2fv)( rmesa->vb.texcoordptr[0] );
+      }
+   }
 
-   if (ind & RADEON_CP_VC_FRMT_ST1) 
-      GL_CALL(MultiTexCoord2fvARB)( GL_TEXTURE1_ARB, rmesa->vb.texcoordptr[1] );
+   if (ind & RADEON_CP_VC_FRMT_ST1) {
+      if (ind & RADEON_CP_VC_FRMT_Q1) {
+         GL_CALL(MultiTexCoord3fvARB)( GL_TEXTURE1, rmesa->vb.texcoordptr[1] );
+      } else {
+         GL_CALL(MultiTexCoord2fvARB)( GL_TEXTURE1, rmesa->vb.texcoordptr[1] );
+      }
+   }
 }
 
 
@@ -468,7 +521,7 @@
 {
    GET_CURRENT_CONTEXT(ctx);
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][15];
+   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
    GLuint i, nrverts;
 
    if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_PRIMS))
@@ -585,29 +638,53 @@
 	    ind |= RADEON_CP_VC_FRMT_N0;
 	 }
       } else {
-	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX0][2] != 0.0F ||
-	     ctx->Current.Attrib[VERT_ATTRIB_TEX0][3] != 1.0) {
+	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX0][3] != 1.0) {
 	    if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-	       fprintf(stderr, "%s: rq0\n", __FUNCTION__);
+	       fprintf(stderr, "%s: q0\n", __FUNCTION__);
 	    return GL_FALSE;
 	 }
 	 ind |= RADEON_CP_VC_FRMT_ST0;
+
+/* FIXME: This doesnt work as expected */
+	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX0][2] != 0.0F)
+	    ind |= RADEON_CP_VC_FRMT_Q0;
+/* workaround? */
+#if 1
+	 if (ctx->Texture.Unit[0]._ReallyEnabled == TEXTURE_CUBE_BIT)
+	    ind |= RADEON_CP_VC_FRMT_Q0;
+#endif
+	 if( RADEON_DEBUG & DEBUG_VFMT)
+	    fprintf( stderr, "%s: ctx->Current.Attrib[VERT_ATTRIB_TEX0][0..3] = %f %f %f %f\n",
+	       __FUNCTION__,
+	       ctx->Current.Attrib[VERT_ATTRIB_TEX0][0],
+	       ctx->Current.Attrib[VERT_ATTRIB_TEX0][1],
+	       ctx->Current.Attrib[VERT_ATTRIB_TEX0][2],
+	       ctx->Current.Attrib[VERT_ATTRIB_TEX0][3]);
       }
    }
 
+#if 0
    if (ctx->Texture.Unit[1]._ReallyEnabled) {
+#else
+   if (1) {
+#endif
       if (ctx->Texture.Unit[1].TexGenEnabled) {
 	 if (rmesa->TexGenNeedNormals[1]) {
 	    ind |= RADEON_CP_VC_FRMT_N0;
 	 }
       } else {
-	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX1][2] != 0.0F ||
-	     ctx->Current.Attrib[VERT_ATTRIB_TEX1][3] != 1.0) {
+	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX1][3] != 1.0) {
 	    if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-	       fprintf(stderr, "%s: rq1\n", __FUNCTION__);
+	       fprintf(stderr, "%s: q1\n", __FUNCTION__);
 	    return GL_FALSE;
 	 }
 	 ind |= RADEON_CP_VC_FRMT_ST1;
+/* this wont work.. */
+	 if (ctx->Current.Attrib[VERT_ATTRIB_TEX1][2] != 0.0F)
+	    ind |= RADEON_CP_VC_FRMT_Q1;
+/* workaround? */
+	 if (ctx->Texture.Unit[1]._ReallyEnabled == TEXTURE_CUBE_BIT)
+	    ind |= RADEON_CP_VC_FRMT_Q1;
       }
    }
 
@@ -674,6 +751,11 @@
       rmesa->vb.vertex_size += 2;
       rmesa->vb.texcoordptr[0][0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0][0];
       rmesa->vb.texcoordptr[0][1] = ctx->Current.Attrib[VERT_ATTRIB_TEX0][1];   
+/* FIXME: first check if Q0 is set from projtex or cubemap/3dtex */
+      if (ind & RADEON_CP_VC_FRMT_Q0) {
+         rmesa->vb.texcoordptr[0][2] = ctx->Current.Attrib[VERT_ATTRIB_TEX0][2];   
+         rmesa->vb.vertex_size += 1;
+      }
    } 
 
    if (ind & RADEON_CP_VC_FRMT_ST1) {
@@ -681,8 +763,13 @@
       rmesa->vb.vertex_size += 2;
       rmesa->vb.texcoordptr[1][0] = ctx->Current.Attrib[VERT_ATTRIB_TEX1][0];
       rmesa->vb.texcoordptr[1][1] = ctx->Current.Attrib[VERT_ATTRIB_TEX1][1];
+      if (ind & RADEON_CP_VC_FRMT_Q1) {
+         rmesa->vb.texcoordptr[1][2] = ctx->Current.Attrib[VERT_ATTRIB_TEX1][2];   
+         rmesa->vb.vertex_size += 1;
+      }
    } 
 
+/* FIXME: also check projformat/installed_projformat ? */
    if (rmesa->vb.installed_vertex_format != rmesa->vb.vertex_format) {
       if (RADEON_DEBUG & DEBUG_VFMT)
 	 fprintf(stderr, "reinstall on vertex_format change\n");
@@ -811,6 +898,7 @@
    /* Need to arrange to save vertices here?  Or always copy from dma (yuk)?
     */
    if (!rmesa->dma.flush) {
+/* FIXME: what are these constants? */
       if (rmesa->dma.current.ptr + 12*rmesa->vb.vertex_size*4 > 
 	  rmesa->dma.current.end) {
 	 RADEON_NEWPRIM( rmesa );
@@ -982,12 +1070,14 @@
    vfmt->EvalMesh2 = radeon_fallback_EvalMesh2;
    vfmt->EvalPoint1 = radeon_fallback_EvalPoint1;
    vfmt->EvalPoint2 = radeon_fallback_EvalPoint2;
-   vfmt->TexCoord3f = radeon_fallback_TexCoord3f;
-   vfmt->TexCoord3fv = radeon_fallback_TexCoord3fv;
    vfmt->TexCoord4f = radeon_fallback_TexCoord4f;
    vfmt->TexCoord4fv = radeon_fallback_TexCoord4fv;
+#if 1	/* \todo: eliminate fallback  */
+   vfmt->TexCoord3f = radeon_fallback_TexCoord3f;
+   vfmt->TexCoord3fv = radeon_fallback_TexCoord3fv;
    vfmt->MultiTexCoord3fARB = radeon_fallback_MultiTexCoord3fARB;
    vfmt->MultiTexCoord3fvARB = radeon_fallback_MultiTexCoord3fvARB;
+#endif
    vfmt->MultiTexCoord4fARB = radeon_fallback_MultiTexCoord4fARB;
    vfmt->MultiTexCoord4fvARB = radeon_fallback_MultiTexCoord4fvARB;
    vfmt->Vertex4f = radeon_fallback_Vertex4f;
@@ -1027,10 +1117,18 @@
    make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
    make_empty_list( &rmesa->vb.dfn_cache.Normal3f );
    make_empty_list( &rmesa->vb.dfn_cache.Normal3fv );
+#if 0
+   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3f );
+   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3fv );
+#endif
    make_empty_list( &rmesa->vb.dfn_cache.TexCoord2f );
    make_empty_list( &rmesa->vb.dfn_cache.TexCoord2fv );
    make_empty_list( &rmesa->vb.dfn_cache.TexCoord1f );
    make_empty_list( &rmesa->vb.dfn_cache.TexCoord1fv );
+#if 0
+   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
+   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
+#endif
    make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
    make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
    make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
@@ -1079,10 +1177,18 @@
    free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
    free_funcs( &rmesa->vb.dfn_cache.Normal3f );
    free_funcs( &rmesa->vb.dfn_cache.Normal3fv );
+#if 0
+   free_funcs( &rmesa->vb.dfn_cache.TexCoord3f );
+   free_funcs( &rmesa->vb.dfn_cache.TexCoord3fv );
+#endif
    free_funcs( &rmesa->vb.dfn_cache.TexCoord2f );
    free_funcs( &rmesa->vb.dfn_cache.TexCoord2fv );
    free_funcs( &rmesa->vb.dfn_cache.TexCoord1f );
    free_funcs( &rmesa->vb.dfn_cache.TexCoord1fv );
+#if 0
+   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
+   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
+#endif
    free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
    free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
    free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h,v
retrieving revision 1.4
diff -u -r1.4 radeon_vtxfmt.h
--- Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h	31 Jan 2005 23:40:06 -0000	1.4
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h	5 Jun 2005 18:19:01 -0000
@@ -49,6 +49,7 @@
 extern void radeonVtxfmtUnbindContext( GLcontext *ctx );
 
 extern void radeon_copy_to_current( GLcontext *ctx );
+extern void VFMT_FALLBACK( const char *caller );
 
 #define DFN( FUNC, CACHE)				\
 do {							\
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c,v
retrieving revision 1.6
diff -u -r1.6 radeon_vtxfmt_c.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c	16 Feb 2005 15:02:25 -0000	1.6
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c	5 Jun 2005 18:19:01 -0000
@@ -510,88 +510,142 @@
 
 /* TexCoord
  */
-static void radeon_TexCoord1f( GLfloat s )
+
+/* MultiTexcoord
+ * 
+ * Technically speaking, these functions should subtract GL_TEXTURE0 from
+ * \c target before masking and using it.  The value of GL_TEXTURE0 is 0x84C0,
+ * which has the low-order 5 bits 0.  For all possible valid values of 
+ * \c target.  Subtracting GL_TEXTURE0 has the net effect of masking \c target
+ * with 0x1F.  Masking with 0x1F and then masking with 0x01 is redundant, so
+ * the subtraction has been omitted.
+ */
+
+/* \todo for tmu3: maybe (target & 2 ? 2 : target & 1) is more save than (target & 2) */
+static void radeon_MultiTexCoord1fARB( GLenum target, GLfloat s  )
 {
    GET_CURRENT_CONTEXT(ctx);
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = 0;
+   GLint unit = (target & 1);
+   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
+
+   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
+   case TEXTURE_CUBE_BIT:
+      dest[2] = 0.0;
+      /* FALLTHROUGH */
+   case TEXTURE_2D_BIT:
+   case TEXTURE_RECT_BIT:
+   case TEXTURE_1D_BIT:
+   default:
+      dest[1] = 0.0;
+      dest[0] = s;
+   }
 }
 
-static void radeon_TexCoord1fv( const GLfloat *v )
+static void radeon_MultiTexCoord2fARB( GLenum target, GLfloat s, GLfloat t )
 {
    GET_CURRENT_CONTEXT(ctx);
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = 0;
+   GLint unit = (target & 1);
+   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
+
+   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
+   case TEXTURE_CUBE_BIT:
+      dest[2] = 0.0;
+      /* FALLTHROUGH */
+   case TEXTURE_2D_BIT:
+   case TEXTURE_RECT_BIT:
+   case TEXTURE_1D_BIT:
+   default:
+      dest[1] = t;
+      dest[0] = s;
+   }
 }
 
-static void radeon_TexCoord2f( GLfloat s, GLfloat t )
+#if 0	/* \todo: make this work */
+static void radeon_MultiTexCoord3fARB(GLenum target, GLfloat s, GLfloat t, GLfloat r)
 {
    GET_CURRENT_CONTEXT(ctx);
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = t;
+   GLint unit = (target & 1);
+   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
+
+   if( RADEON_DEBUG & DEBUG_VFMT)
+      fprintf( stderr, "%s: rmesa->vb.texcoordptr[%d]=0x%0x\n", __FUNCTION__, unit, rmesa->vb.texcoordptr[unit]);
+   if( RADEON_DEBUG & DEBUG_VFMT)
+      fprintf( stderr, "%s: rmesa->vb.vertex_size=%d\n", __FUNCTION__, rmesa->vb.vertex_size);
+   if( RADEON_DEBUG & DEBUG_VFMT)
+      fprintf( stderr, "%s: old str: %f %f %f\n", __FUNCTION__, dest[0], dest[1], dest[2]);
+
+   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
+   case TEXTURE_CUBE_BIT:
+      dest[2] = r;
+      dest[1] = t;
+      dest[0] = s;
+      if( (RADEON_DEBUG & DEBUG_VFMT) && (RADEON_DEBUG & DEBUG_VERBOSE))
+         fprintf( stderr, "%s: emit s %f  t %f  r %f to dest\n",
+            __FUNCTION__, s, t, r);
+         
+      break;
+   default:
+      VFMT_FALLBACK(__FUNCTION__);
+      GL_CALL(MultiTexCoord3fARB)(target, s, t, r);
+      return;	
+   }
 }
+#endif
 
-static void radeon_TexCoord2fv( const GLfloat *v )
+static void radeon_TexCoord1f(GLfloat s)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = v[1];
+   radeon_MultiTexCoord1fARB(GL_TEXTURE0, s);
 }
 
+static void radeon_TexCoord2f(GLfloat s, GLfloat t)
+{
+   radeon_MultiTexCoord2fARB(GL_TEXTURE0, s, t);
+}
 
-/* MultiTexcoord
- * 
- * Technically speaking, these functions should subtract GL_TEXTURE0 from
- * \c target before masking and using it.  The value of GL_TEXTURE0 is 0x84C0,
- * which has the low-order 5 bits 0.  For all possible valid values of 
- * \c target.  Subtracting GL_TEXTURE0 has the net effect of masking \c target
- * with 0x1F.  Masking with 0x1F and then masking with 0x01 is redundant, so
- * the subtraction has been omitted.
- */
+#if 0
+static void radeon_TexCoord3f(GLfloat s, GLfloat t, GLfloat r)
+{
+   radeon_MultiTexCoord3fARB(GL_TEXTURE0, s, t, r);
+}
+#endif
 
-static void radeon_MultiTexCoord1fARB( GLenum target, GLfloat s  )
+static void radeon_TexCoord1fv(const GLfloat *v)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 1];
-   dest[0] = s;
-   dest[1] = 0;
+   radeon_MultiTexCoord1fARB(GL_TEXTURE0, v[0]);
 }
 
-static void radeon_MultiTexCoord1fvARB( GLenum target, const GLfloat *v )
+static void radeon_TexCoord2fv(const GLfloat *v)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 1];
-   dest[0] = v[0];
-   dest[1] = 0;
+   radeon_MultiTexCoord2fARB(GL_TEXTURE0, v[0], v[1]);
 }
 
-static void radeon_MultiTexCoord2fARB( GLenum target, GLfloat s, GLfloat t )
+#if 0
+static void radeon_TexCoord3fv(const GLfloat *v)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 1];
-   dest[0] = s;
-   dest[1] = t;
+   radeon_MultiTexCoord3fARB(GL_TEXTURE0, v[0], v[1], v[2]);
 }
+#endif
 
-static void radeon_MultiTexCoord2fvARB( GLenum target, const GLfloat *v )
+static void radeon_MultiTexCoord1fvARB(GLenum target, const GLfloat *v)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 1];
-   dest[0] = v[0];
-   dest[1] = v[1];
+   radeon_MultiTexCoord1fARB(target, v[0]);
+}
+
+static void radeon_MultiTexCoord2fvARB(GLenum target, const GLfloat *v)
+{
+   radeon_MultiTexCoord2fARB(target, v[0], v[1]);
+}
+
+#if 0
+static void radeon_MultiTexCoord3fvARB(GLenum target, const GLfloat *v)
+{
+   radeon_MultiTexCoord3fARB(target, v[0], v[1], v[2]);
 }
+#endif
+
 
 static struct dynfn *lookup( struct dynfn *l, int key )
 {
@@ -615,7 +669,8 @@
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
    int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
    struct dynfn *dfn;							\
-									\
+if( RADEON_DEBUG & DEBUG_VFMT) fprintf( stderr, "%s: mask:%0x active:%0x\n", \
+__FUNCTION__ , MASK, ACTIVE);									\
    dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
    if (dfn == 0)							\
       dfn = rmesa->vb.codegen.FN( ctx, key );				\
@@ -746,6 +801,10 @@
 #define ACTIVE_ST1 RADEON_CP_VC_FRMT_ST1
 #define ACTIVE_ST_ALL (RADEON_CP_VC_FRMT_ST1|RADEON_CP_VC_FRMT_ST0)
 
+#define ACTIVE_STQ0 (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_Q0)
+#define ACTIVE_STQ1 (RADEON_CP_VC_FRMT_ST1|RADEON_CP_VC_FRMT_Q1)
+#define ACTIVE_STQ_ALL (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_Q0|RADEON_CP_VC_FRMT_ST1|RADEON_CP_VC_FRMT_Q1)
+
 /* Each codegen function should be able to be fully specified by a
  * subsetted version of rmesa->vb.vertex_format.
  */
@@ -755,13 +814,19 @@
 #define MASK_ST0     (MASK_SPEC|ACTIVE_SPEC)
 #define MASK_ST1     (MASK_ST0|ACTIVE_ST0)
 #define MASK_ST_ALL  (MASK_ST1|ACTIVE_ST1)
-#define MASK_VERTEX  (MASK_ST_ALL|ACTIVE_FPALPHA) 
+
+#define MASK_STQ0    (MASK_ST0)
+#define MASK_STQ1    (MASK_ST_ALL|ACTIVE_STQ1)
+#define MASK_STQ_ALL (MASK_ST_ALL|ACTIVE_STQ_ALL)
+
+#define MASK_VERTEX  (MASK_STQ_ALL|ACTIVE_FPALPHA) 
 
 
 typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
 typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
 typedef void (*p2f)( GLfloat, GLfloat );
 typedef void (*p1f)( GLfloat );
+typedef void (*pe3f)( GLenum, GLfloat, GLfloat, GLfloat );
 typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
 typedef void (*pe1f)( GLenum, GLfloat );
 typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
@@ -808,6 +873,12 @@
 CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, ACTIVE_SPEC,
 	(const GLfloat *v), (v))
 
+#if 0
+CHOOSE(TexCoord3f, p3f, MASK_STQ0, ACTIVE_STQ0, 
+       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
+CHOOSE(TexCoord3fv, pfv, MASK_STQ0, ACTIVE_STQ0, 
+       (const GLfloat *v), (v))
+#endif
 CHOOSE(TexCoord2f, p2f, MASK_ST0, ACTIVE_ST0, 
        (GLfloat a,GLfloat b), (a,b))
 CHOOSE(TexCoord2fv, pfv, MASK_ST0, ACTIVE_ST0, 
@@ -817,6 +888,12 @@
 CHOOSE(TexCoord1fv, pfv, MASK_ST0, ACTIVE_ST0, 
        (const GLfloat *v), (v))
 
+#if 0
+CHOOSE(MultiTexCoord3fARB, pe3f, MASK_STQ_ALL, ACTIVE_STQ_ALL,
+	 (GLenum u,GLfloat a,GLfloat b,GLfloat c), (u,a,b,c))
+CHOOSE(MultiTexCoord3fvARB, pefv, MASK_STQ_ALL, ACTIVE_STQ_ALL,
+	(GLenum u,const GLfloat *v), (u,v))
+#endif
 CHOOSE(MultiTexCoord2fARB, pe2f, MASK_ST_ALL, ACTIVE_ST_ALL,
 	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
 CHOOSE(MultiTexCoord2fvARB, pefv, MASK_ST_ALL, ACTIVE_ST_ALL,
@@ -851,12 +928,20 @@
    vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
    vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
    vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
+#if 0
+   vfmt->MultiTexCoord3fARB = choose_MultiTexCoord3fARB;
+   vfmt->MultiTexCoord3fvARB = choose_MultiTexCoord3fvARB;
+#endif
    vfmt->Normal3f = choose_Normal3f;
    vfmt->Normal3fv = choose_Normal3fv;
    vfmt->TexCoord1f = choose_TexCoord1f;
    vfmt->TexCoord1fv = choose_TexCoord1fv;
    vfmt->TexCoord2f = choose_TexCoord2f;
    vfmt->TexCoord2fv = choose_TexCoord2fv;
+#if 0
+   vfmt->TexCoord3f = choose_TexCoord3f;
+   vfmt->TexCoord3fv = choose_TexCoord3fv;
+#endif
    vfmt->Vertex2f = choose_Vertex2f;
    vfmt->Vertex2fv = choose_Vertex2fv;
    vfmt->Vertex3f = choose_Vertex3f;
@@ -887,10 +972,25 @@
    gen->Color4ubv = codegen_noop;
    gen->Normal3f = codegen_noop;
    gen->Normal3fv = codegen_noop;
+
+#if 0
+   gen->TexCoord3f = codegen_noop;
+   gen->TexCoord3fv = codegen_noop;
+#endif
    gen->TexCoord2f = codegen_noop;
    gen->TexCoord2fv = codegen_noop;
+   gen->TexCoord1f = codegen_noop;
+   gen->TexCoord1fv = codegen_noop;
+
+#if 0
+   gen->MultiTexCoord3fARB = codegen_noop;
+   gen->MultiTexCoord3fvARB = codegen_noop;
+#endif
    gen->MultiTexCoord2fARB = codegen_noop;
    gen->MultiTexCoord2fvARB = codegen_noop;
+   gen->MultiTexCoord1fARB = codegen_noop;
+   gen->MultiTexCoord1fvARB = codegen_noop;
+
    gen->Vertex2f = codegen_noop;
    gen->Vertex2fv = codegen_noop;
    gen->Color3ub = codegen_noop;
@@ -903,10 +1003,6 @@
    gen->SecondaryColor3fvEXT = codegen_noop;
    gen->SecondaryColor3ubEXT = codegen_noop;
    gen->SecondaryColor3ubvEXT = codegen_noop;
-   gen->TexCoord1f = codegen_noop;
-   gen->TexCoord1fv = codegen_noop;
-   gen->MultiTexCoord1fARB = codegen_noop;
-   gen->MultiTexCoord1fvARB = codegen_noop;
 
    if (useCodegen) {
 #if defined(USE_X86_ASM)
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c,v
retrieving revision 1.5
diff -u -r1.5 radeon_vtxfmt_sse.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c	16 Feb 2005 15:02:25 -0000	1.5
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c	5 Jun 2005 18:19:01 -0000
@@ -154,6 +154,7 @@
    }
 }
 
+#if 0 /* from r200: Temporarily disabled as it is broken w/the new cubemap code. - idr */
 static struct dynfn *radeon_makeSSETexCoord2fv( GLcontext *ctx, int key )
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
@@ -207,6 +208,7 @@
    }
    return dfn;
 }
+#endif
 
 void radeonInitSSECodegen( struct dfn_generators *gen )
 {
@@ -215,10 +217,12 @@
       gen->Normal3f = (void *) radeon_makeSSENormal3f;
       gen->Color3fv = (void *) radeon_makeSSEColor3fv;
       gen->Color3f = (void *) radeon_makeSSEColor3f;
+#if 0 /* from r200: Temporarily disabled as it is broken w/the new cubemap code. - idr */
       gen->TexCoord2fv = (void *) radeon_makeSSETexCoord2fv;
       gen->TexCoord2f = (void *) radeon_makeSSETexCoord2f;
       gen->MultiTexCoord2fvARB = (void *) radeon_makeSSEMultiTexCoord2fv;
       gen->MultiTexCoord2fARB = (void *) radeon_makeSSEMultiTexCoord2f;
+#endif
    }
 }
 
Index: Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c
===================================================================
RCS file: /cvs/mesa/Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c,v
retrieving revision 1.4
diff -u -r1.4 radeon_vtxfmt_x86.c
--- Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c	16 Feb 2005 15:02:25 -0000	1.4
+++ Mesa/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c	5 Jun 2005 18:19:01 -0000
@@ -335,6 +335,7 @@
 
 
 
+#if 0 /* from r200: Temporarily disabled as it is broken w/the new cubemap code. - idr */
 struct dynfn *radeon_makeX86TexCoord2fv( GLcontext *ctx, int key )
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
@@ -395,6 +396,7 @@
    }      
    return dfn;
 }
+#endif
 
 
 void radeonInitX86Codegen( struct dfn_generators *gen )
@@ -405,10 +407,12 @@
    gen->Color4ubv = radeon_makeX86Color4ubv; /* PKCOLOR only */
    gen->Normal3f = radeon_makeX86Normal3f;
    gen->Normal3fv = radeon_makeX86Normal3fv;
+#if 0 /* from r200: Temporarily disabled as it is broken w/the new cubemap code. - idr */
    gen->TexCoord2f = radeon_makeX86TexCoord2f;
    gen->TexCoord2fv = radeon_makeX86TexCoord2fv;
    gen->MultiTexCoord2fARB = radeon_makeX86MultiTexCoord2fARB;
    gen->MultiTexCoord2fvARB = radeon_makeX86MultiTexCoord2fvARB;
+#endif
    gen->Color3f = radeon_makeX86Color3f;
    gen->Color3fv = radeon_makeX86Color3fv;