(#823): Fix geometry computation for individual layers

(#823) correctly points out that commit
e3bb911b2cd860e4e4671c87ea9712913bb35d26 broke geometry computation.
I didn't see how it could, since it didn't even have to do with
extracting bounding boxes from elements.

It turns out that I forgot to copy the DrawingCtx.drawsub_stack for
the nested context.  So, for an SVG with structure

  a
    b
      c -> with isolation context (due to non-100% opacity / etc.)
        d
        e
        f id="draw_only_this_one"
        g

if one calls geometry_for_layer("draw_only_this_one"), draw_tree() would
properly create a drawsub_stack like (a b c f), but then the nested
rendering would go:

  stack_top=a

  stack_top=b

  stack_top=c (creates isolation context, and a nested DrawingCtx)

  stack_top=empty (nested DrawingCtx has an empty drawsub_stack) ->
    draw every child of c since there's no drawsub_stack to limit it,
    instead of just drawing f in there.

https://gitlab.gnome.org/GNOME/librsvg/-/issues/823

Part-of: <https://gitlab.gnome.org/GNOME/librsvg/-/merge_requests/697>

1 files changed, 1 insertions, 1 deletions

diff --git a/src/drawing_ctx.rs b/src/drawing_ctx.rs
index 41029551..cde3c61d 100644
--- a/src/drawing_ctx.rs
+++ b/src/drawing_ctx.rs
@@ -326,7 +326,7 @@ impl DrawingCtx {
             cr,
             user_language: self.user_language.clone(),
             viewport_stack: self.viewport_stack.clone(),
-            drawsub_stack: Vec::new(),
+            drawsub_stack: self.drawsub_stack.clone(),
             measuring: self.measuring,
             testing: self.testing,
         }