bug#44655: 28.0.50; Oversized SVG margin

[Paul Eggert]

On 11/6/21 05:55, Alan Third wrote:

> There seems to be a lot of churn in the librsvg API at the moment, and
> their documentation isn't keeping up as it still doesn't mark
> rsvg_handle_get_dimensions as deprecated.
>
> It appears they've introduced rsvg_handle_get_intrinsic_size_in_pixels
> which does the same thing (but better) than my sizing code, but anyone
> using librsvg 2.46-2.52, which is probably most people for now, can't
> use it and probably don't want to use the old scheme, so I expect
> there's no point replacing my code at the moment.

If it improves on your code then let's try using it for bleeding-edge 
librsvg (2.52.0+).


> Perhaps what we should do is move the final "else" section of the
> previous code block (where rsvg_handle_get_geometry_for_layer is
> called) into its own block which is executed
>
>    if (viewbox_width == 0 || viewbox_height == 0)
>
> instead of only in the case where rsvg_handle_get_intrinsic_dimensions
> fails to return any dimensions. That way we should have *some*
> dimensions without having to call rsvg_handle_get_dimensions.

I installed a patch into master to do that, along with other patches to 
try using svg_handle_get_intrinsic_size_in_pixels with bleeding-edge 
librsvg, and to catch some potential integer overflow problems I noticed 
while doing all this (see attached).

0001-More-robust-svg_load_image-fallback.patch
Description: Text Data

0002-Simplify-svg_load_image.patch
Description: Text Data

0003-Improve-overflow-checking-in-svg_load_image.patch
Description: Text Data

0004-Prefer-rsvg_handle_get_intrinsic_size_in_pixels.patch
Description: Text Data