tiger.svg not rendering correctly

[JaimeIvanCervantes]

Hello, I was able to render camera.svg and other 3 simple svgs correctly using the example on tests/render/main.cpp but tiger.svg is failing to render properly:

Here's my result when rendering the file camera.svg:

And here's my result when rendering the file tiger.svg:

I use the exact same code for both, the only thing that I changed was the name of the file. I wonder if this is just happening to me, or if you also are seeing this issue too?

[saeitsystems]

In my application svgren renders the tiger.svg without problems:

(under windows with VisualStudio 2015)

[igagis]

I also don't see problems with tiger.svg, at least I tried recently under Windows and under Linux. @JaimeIvanCervantes on which OS are you trying?

[saeitsystems]

I got a different tiger.svg from wikipedia which causes an assertion in svgren:

See attached tiger.svg image tiger.zip

[igagis]

Hm, interesting, let me try...

[igagis]

Ok, it uses Cubuc_Smooth_Rel elements in some path which is not implemented in svgren. I need to implement it.

[saeitsystems]

It is not urgent to me but since it is a standard file on wikipmedia it would be better not to have an assertion ... https://commons.wikimedia.org/wiki/File:Ghostscript_Tiger.svg

[igagis]

svgren is not a complete support of the SVG standard yet, as the standard is quite big, so I implement features on demand. So thank you for finding missing stuff and providing sample SVGs for those, I add all such SVGs to the test data set. Assertions are only active in Debug configuration, in Release there are no assertions done, but I checked that tiger.svg in Relese mode, it looks quite distorted :)

[saeitsystems]

In release mode I got this result:

[igagis]

Yep, it's quite distorted because of those missing path steps. I'm trying to fix it now.

[saeitsystems]

Interesting is that the new tiger.svg from mediawiki looks identical but the file is about 30% smaller than your test tiger.svg

[igagis]

@saeitsystems obviously, it is because in my tiger.svg version smooth curve steps are converted to non-smooth ones, which are basically identical, only that smooth ones are shorter in writing. You can see more about smooth/non-smooth cubic curve path steps here https://www.w3.org/TR/SVG/paths.html#PathDataCubicBezierCommands

[saeitsystems]

@igagis that looks quite complicated .... I think your library is already doing a very good job! Are there any favorite SVG editors you would recomend? I use AffinityDesigner on Windows or Mac. I tried Inkscape - I didn't like it because it is slow, the GUI is complicated and not standard windows and there is no UHD support.

[igagis]

Well, I'm not drawing much SVGs on my own, but when I need to check something or edit some SVG I use Inkscape. Because I mainly use Linux envirnoment, it is easy to install it there and it is quite sufficient for small SVG tweaks. I did not search for better editors, so, no idea what else is available.

[JaimeIvanCervantes]

I am using macOS Sierra with clang-802.0.42 and cairo 1.14.10 (using Homebrew), and I'm saving the results to png with the changes on my PR: https://github.com/igagis/svgren/pull/17

camera.svg and other svgs were rendered correctly, it's only tiger.svg that doesn't work. Hmm maybe it's something on my side, I'll keep investigating.

[igagis]

Ok, I think I have also seen some strange results on MacOS, but not that big as your tiger.svg. I will try to reproduce on MacOS.

[igagis]

I reproduced the problem with tiger.svg on MacOS, same as @JaimeIvanCervantes observes, investigating...

[igagis]

For some reason, on MacOS the std lib works differently than on other systems. The problem which causes the tiger.svg bug is that on MacOS the following assertion fails:

std::istringstream s("0c");
float d;
s >> d;
assert(!s.fail())

any idea how to workaround this?

[JaimeIvanCervantes]

I am reproducing the failed assertion:

#include <iostream>
#include <sstream>
#include <assert.h>

using namespace std;

int main() {
        std::istringstream s("0c");
        float d;
        s >> d;

        cout<<"d="<<d<<endl;

        if (s.fail()) {
                cout<<"Error"<<endl;
        }

        assert(!s.fail());

        return 0;
}

And here's the output:

d=0
Error
Assertion failed: (!s.fail()), function main, file istreamstringError.cpp, line 18.
Abort trap: 6

Thinking about best way to get around it.

[JaimeIvanCervantes]

I think the c next to the 0 is causing the failure. When I use std::istringstream s("0") I don't get any errors.

[igagis]

Sure, but the problem is that exactly this "0c" is a valid part of SVG document which is to be parsed as number 0 and 'c' should be ignored. I'm now implementing some workaround for this.

[igagis]

@JaimeIvanCervantes Ok, the problem was actually in svgdom. I made a workaround. The svgdom version 0.2.36 is coming soon, here is the travis build: https://travis-ci.org/igagis/svgdom/builds when it is done, you can upgrade svgdom via Homebrew.

[JaimeIvanCervantes]

Excellent thanks for the quick fix! I actually have my own cmake file with svgdom and svgren dependencies, so I just pulled your changes and it works like a charm, tiger.svg is rendering correctly now ;)

[igagis]

There is still other issue with Wikipedia tiger.svg, so reopening this bug for now, though the fix will come soon.

[igagis]

@saeitsystems Ok, new packages released to nuget: svgdom version 0.2.40 svgren version 0.4.20 please try and close the bug.

[saeitsystems]

The Tiger from wikimedia renders fine now

It seems, I can't close the bug because I didn't open ist

[igagis]

Good, and Jaime also reported that he had no problems on MacOS now. So, closing the bug. Thank you!