Hi,
I've run the SVG 1.1 test suite with the latest version of trunk. Compared with 0.48, 21 tests newly pass and an additional 13 are closer to passing. There are 14 tests that no longer pass, mostly due to the feImage filter primitive no longer being able to load PNG and JPG files. (This was broken in commit 12939.) And there is one test that caused an infinite loop: struct-use-08-b.svg
The test results can be seen at:
http://tavmjong.free.fr/INKSCAPE/W3C_SVG_0.91/htmlInkscapeHarness/index.html
There is an HTML page for each test showing the raw SVG file, and PNG results for Inkscape 0.48, Inkscape 0.49 (a snapshot from some months ago), current trunk (0.91), and the official reference PNG. The formatting isn't the best... sorry. I have not removed tests that Inkscape cannot pass due to animation or the use of the DOM.
Here is a full list of the test results:
New Failures
0.48 -> 0.49
coords-units-01-b.svg filters-conv-01-f.svg (edge effect) filters-displace-01-f.svg painting-marker-02-f.svg (fixed in 0.49 -> 0.91) struct-cond-01-t.svg (fixed in 0.49 -> 0.91) text-tref-02-b.svg (partial)
0.49 -> 0.91
coords-units-01-b.svg filters-composite-04-f.svg filters-displace-01-f.svg filters-image-01-b.svg filters-image-02-b.svg filters-image-03-f.svg filters-image-04-f.svg filters-image-05-f.svg filters-light-02-f.svg struct-use-08-b.svg (*** Infinite loop***)
New Passes
0.48 -> 0.49
filters-background-01-f.svg filters-composite-03-f.svg (partial) filters-comptran-01-b.svg filters-conv-02-f.svg filters-displace-02-f.svg filters-example-01-b.svg filters-felem-01-b.svg (partial) filters-felem-02-f.svg (partial) filters-gauss-02-f.svg filters-gauss-03-f.svg filters-image-03-f.svg filters-image-04-f.svg filters-turb-01-f.svg masking-path-05-f.svg masking-path-07-b.svg (partial) painting-marker-07-f.svg (partial) painting-stroke-06-t.svg paths-data-10-t.svg paths-data-18-f.svg (partial) pservers-grad-17-b.svg (partial) pservers-grad-20-b.svg (partial) struct-frag-05-t.svg struct-image-02-b.svg (partial) text-altglyph-01-b.svg text-deco-01-b.svg text-text-03-b.svg
0.49 -> 0.91
filters-felem-02-f.svg (partial) painting-marker-02-f.svg (partial) struct-image-05-b.svg struct-symbol-01-b.svg struct-use-04-b.svg (partial) struct-use-05-b.svg (partial) struct-use-11-f.svg text-tref-01-b.svg
Tav
I've fixed a handful of test due to the feImage bug. Fixed test marked below. Test results updated.
On Sun, 2014-03-02 at 20:58 +0100, Tavmjong Bah wrote:
Hi,
I've run the SVG 1.1 test suite with the latest version of trunk. Compared with 0.48, 21 tests newly pass and an additional 13 are closer to passing. There are 14 tests that no longer pass, mostly due to the feImage filter primitive no longer being able to load PNG and JPG files. (This was broken in commit 12939.) And there is one test that caused an infinite loop: struct-use-08-b.svg
The test results can be seen at:
http://tavmjong.free.fr/INKSCAPE/W3C_SVG_0.91/htmlInkscapeHarness/index.html
There is an HTML page for each test showing the raw SVG file, and PNG results for Inkscape 0.48, Inkscape 0.49 (a snapshot from some months ago), current trunk (0.91), and the official reference PNG. The formatting isn't the best... sorry. I have not removed tests that Inkscape cannot pass due to animation or the use of the DOM.
Here is a full list of the test results:
New Failures
0.48 -> 0.49
coords-units-01-b.svg filters-conv-01-f.svg (edge effect) filters-displace-01-f.svg painting-marker-02-f.svg (fixed in 0.49 -> 0.91) struct-cond-01-t.svg (fixed in 0.49 -> 0.91) text-tref-02-b.svg (partial)
0.49 -> 0.91
coords-units-01-b.svg filters-composite-04-f.svg FIXED filters-displace-01-f.svg PARTIALLY FIXED filters-image-01-b.svg FIXED filters-image-02-b.svg FIXED (Animation) filters-image-03-f.svg FIXED filters-image-04-f.svg FIXED filters-image-05-f.svg FIXED filters-light-02-f.svg struct-use-08-b.svg (*** Infinite loop***)
New Passes
0.48 -> 0.49
filters-background-01-f.svg filters-composite-03-f.svg (partial) filters-comptran-01-b.svg filters-conv-02-f.svg filters-displace-02-f.svg filters-example-01-b.svg filters-felem-01-b.svg (partial) filters-felem-02-f.svg (partial) filters-gauss-02-f.svg filters-gauss-03-f.svg filters-image-03-f.svg filters-image-04-f.svg filters-turb-01-f.svg masking-path-05-f.svg masking-path-07-b.svg (partial) painting-marker-07-f.svg (partial) painting-stroke-06-t.svg paths-data-10-t.svg paths-data-18-f.svg (partial) pservers-grad-17-b.svg (partial) pservers-grad-20-b.svg (partial) struct-frag-05-t.svg struct-image-02-b.svg (partial) text-altglyph-01-b.svg text-deco-01-b.svg text-text-03-b.svg
0.49 -> 0.91
filters-felem-02-f.svg (partial) painting-marker-02-f.svg (partial) struct-image-05-b.svg struct-symbol-01-b.svg struct-use-04-b.svg (partial) struct-use-05-b.svg (partial) struct-use-11-f.svg text-tref-01-b.svg
Tav
On Mon, 2014-03-03 at 12:26 +0100, Tavmjong Bah wrote:
struct-use-08-b.svg (*** Infinite loop***)
This is actually caused by the removal of the path normalisation code. It's having trouble matching include-bug/svg/../images/struct-use-08-b-cycles.svg to include-bug/images/../svg/../images/struct-use-08-b-cycles.svg
and that path gets really big. So not just an infinite loop, but it'll chomp through your memory quite quickly as it loads an infinite of documents.
I'm going to try and put back some normalisation without using glib; either with boost or with just std::string actions to rebuild the path.
Martin,
On 2014-03-02 20:58 +0100, Tavmjong Bah wrote:
I've run the SVG 1.1 test suite with the latest version of trunk. Compared with 0.48, 21 tests newly pass and an additional 13 are closer to passing. There are 14 tests that no longer pass, mostly due to the feImage filter primitive no longer being able to load PNG and JPG files. (This was broken in commit 12939.) And there is one test that caused an infinite loop: struct-use-08-b.svg
The test results can be seen at:
http://tavmjong.free.fr/INKSCAPE/W3C_SVG_0.91/htmlInkscapeHarness/index.html
There is an HTML page for each test showing the raw SVG file, and PNG results for Inkscape 0.48, Inkscape 0.49 (a snapshot from some months ago), current trunk (0.91), and the official reference PNG. The formatting isn't the best... sorry. I have not removed tests that Inkscape cannot pass due to animation or the use of the DOM.
Here is a full list of the test results:
New Failures
0.48 -> 0.49
coords-units-01-b.svg
The recently committed fix for this failure
- Rev 13111: Fix for W3C test suite coords-units-01-b.svg. Fixes pattern positioning http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/13111
changes how patterns are rendered and edited in Inkscape (affects existing documents and newly created documents with stock and custom patterns):
- A pattern fill is only visible if the object's pattern handles are positioned over the area of the initial tile, at the SVG origin. - Moving an object with a pattern as paint server and default settings moves the pattern handles as well -> the pattern disappears. - Scaling via pattern handles scales the content within the original tile (clipped by the tile's viewbox (?)), not the tile size itself as before. - Rotating a pattern via pattern handles rotates the content within the original tile (clipped by the viewbox (?)), and not the pattern fill itself as before. - If the pattern handles are positioned exactly at the position of the original object(s), the pattern may render invisible (it just disappears).
As result of the above changes, many patterns either don't render a visible content, or - as e.g. with most of the stripes patterns - as solid fill (due to scaling & clipping within the original tile AFAIU).
On the plus side, it fixes this bug: - Bug #1286407 tiled sections of circles in pattern https://bugs.launchpad.net/inkscape/+bug/1286407
filters-conv-01-f.svg (edge effect) filters-displace-01-f.svg painting-marker-02-f.svg (fixed in 0.49 -> 0.91) struct-cond-01-t.svg (fixed in 0.49 -> 0.91) text-tref-02-b.svg (partial)
0.49 -> 0.91
coords-units-01-b.svg filters-composite-04-f.svg filters-displace-01-f.svg filters-image-01-b.svg filters-image-02-b.svg filters-image-03-f.svg filters-image-04-f.svg filters-image-05-f.svg filters-light-02-f.svg struct-use-08-b.svg (*** Infinite loop***)
New Passes
0.48 -> 0.49
filters-background-01-f.svg filters-composite-03-f.svg (partial) filters-comptran-01-b.svg filters-conv-02-f.svg filters-displace-02-f.svg filters-example-01-b.svg filters-felem-01-b.svg (partial) filters-felem-02-f.svg (partial) filters-gauss-02-f.svg filters-gauss-03-f.svg filters-image-03-f.svg filters-image-04-f.svg filters-turb-01-f.svg masking-path-05-f.svg masking-path-07-b.svg (partial) painting-marker-07-f.svg (partial) painting-stroke-06-t.svg paths-data-10-t.svg paths-data-18-f.svg (partial) pservers-grad-17-b.svg (partial) pservers-grad-20-b.svg (partial) struct-frag-05-t.svg struct-image-02-b.svg (partial) text-altglyph-01-b.svg text-deco-01-b.svg text-text-03-b.svg
0.49 -> 0.91
filters-felem-02-f.svg (partial) painting-marker-02-f.svg (partial) struct-image-05-b.svg struct-symbol-01-b.svg struct-use-04-b.svg (partial) struct-use-05-b.svg (partial) struct-use-11-f.svg text-tref-01-b.svg
On Wed, 2014-03-05 at 21:50 +0100, su_v wrote:
On 2014-03-02 20:58 +0100, Tavmjong Bah wrote:
I've run the SVG 1.1 test suite with the latest version of trunk. Compared with 0.48, 21 tests newly pass and an additional 13 are closer to passing. There are 14 tests that no longer pass, mostly due to the feImage filter primitive no longer being able to load PNG and JPG files. (This was broken in commit 12939.) And there is one test that caused an infinite loop: struct-use-08-b.svg
The test results can be seen at:
http://tavmjong.free.fr/INKSCAPE/W3C_SVG_0.91/htmlInkscapeHarness/index.html
There is an HTML page for each test showing the raw SVG file, and PNG results for Inkscape 0.48, Inkscape 0.49 (a snapshot from some months ago), current trunk (0.91), and the official reference PNG. The formatting isn't the best... sorry. I have not removed tests that Inkscape cannot pass due to animation or the use of the DOM.
Here is a full list of the test results:
New Failures
0.48 -> 0.49
coords-units-01-b.svg
The recently committed fix for this failure
- Rev 13111: Fix for W3C test suite coords-units-01-b.svg. Fixes pattern positioning http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/13111
changes how patterns are rendered and edited in Inkscape (affects existing documents and newly created documents with stock and custom patterns):
- A pattern fill is only visible if the object's pattern handles are
positioned over the area of the initial tile, at the SVG origin.
- Moving an object with a pattern as paint server and default settings
moves the pattern handles as well -> the pattern disappears.
- Scaling via pattern handles scales the content within the original
tile (clipped by the tile's viewbox (?)), not the tile size itself as before.
- Rotating a pattern via pattern handles rotates the content within the
original tile (clipped by the viewbox (?)), and not the pattern fill itself as before.
- If the pattern handles are positioned exactly at the position of the
original object(s), the pattern may render invisible (it just disappears).
As result of the above changes, many patterns either don't render a visible content, or - as e.g. with most of the stripes patterns - as solid fill (due to scaling & clipping within the original tile AFAIU).
On the plus side, it fixes this bug:
- Bug #1286407 tiled sections of circles in pattern https://bugs.launchpad.net/inkscape/+bug/1286407
I have a fix... needs to be cleaned up....
Tav
participants (3)
-
Martin Owens -
su_v -
Tavmjong Bah