mesa: fix a render to texture FBO validation bug

When glTexImage() is called we need to re-validate any FBOs that point to
the texture (i.e. render-to-texture) since changing the texture's size/format
will effect FBO completeness.

We don't keep a list of all FBOs rendering into each texture (which would be
a bit messy) so we check all FBOs in existance.  To optimize this, the
gl_texture_object->_RenderToTexture flag is used to avoid checking textures
that have never been used as renderbuffers.  So, we only walk over all FBOs
(there's usually only a few) when glTexImage() modifies a RTT texture.

Fixes a bug seen in shadowtex.c when toggling packed depth/stencil mode.

1 files changed, 9 insertions, 0 deletions

diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index 1a191cd288c..eec8f1564b6 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -1469,6 +1469,15 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target,
    if (texObj) {
       _mesa_set_texture_attachment(ctx, fb, att, texObj, textarget,
                                    level, zoffset);
+      /* Set the render-to-texture flag.  We'll check this flag in
+       * glTexImage() and friends to determine if we need to revalidate
+       * any FBOs that might be rendering into this texture.
+       * This flag never gets cleared since it's non-trivial to determine
+       * when all FBOs might be done rendering to this texture.  That's OK
+       * though since it's uncommon to render to a texture then repeatedly
+       * call glTexImage() to change images in the texture.
+       */
+      texObj->_RenderToTexture = GL_TRUE;
    }
    else {
       _mesa_remove_attachment(ctx, att);