Bug 52274 – [GTK] Some SVG images fail to load with a "Document is empty" error

• Component: WebKitGTK
• Status: RESOLVED INVALID

[Bakhtiar Hasmanan]

Created attachment 78653 [details]
GTKLauncher SVG test page

I use the patch and source available at http://ftp5.gwdg.de/pub/opensuse/repositories/windows:/mingw:/win32/SLE_11/src/

My setup:
TDM MinGW 4.4.1+MSYS (not crosscompile)
gperf 3.0.1 msys (old) from http://www.foyeh.org/msys/
bison 2.4.2 msys
flex 2.5.35 msys
ActivePerl 5.12
GTK-2.17.2
Glib-2.22.5
libsoup-2.32.2
gstreamer-0.10.31
cairo 1.8.10
pango-1.28.0
enchant-1.6.0
libxslt-1.1.26
libxml2-2.7.7
sqlite-3.7.3

It compiled fine but when I test svg rendering, gtklauncher give error like in attachment and the console prints:
(GtkLauncher.exe:4080): Gtk-CRITICAL **: gtk_container_add: assertion `GTK_IS_CONTAINER (container)' failed
on each failed svg images but when render svg as image it works (or kind of)

Which component that failed in above testcase?
OpenSuSe crosscompile's binary finished the svg tests well. All sources/devlibs in my build came from OpenSuSe except that I need to avoid GTK >=2.18

[Bakhtiar Hasmanan]

Created attachment 78654 [details]
Configure log

[Martin Robinson]

Do you think you could run GtkLauncher with --g-fatal-warnings and get a stack trace for the critical warning?

[Bakhtiar Hasmanan]

(In reply to comment #2)
> Do you think you could run GtkLauncher with --g-fatal-warnings and get a stack trace for the critical warning?

With that argument, GTKLauncher quit after show a dialog box for the same error above.

[Martin Robinson]

At first, I didn't realize what the issue here was. You didn't link to the test page anywhere. The problem here is that some of the SVG images on the page http://a.deveria.com/svg/css_svg_test.html fail to load with an XML parsing error "Document is empty." I believe I've seen errors like these before, but I'm not sure if there is a bug open for it.

[Bakhtiar Hasmanan]

(In reply to comment #4)
> At first, I didn't realize what the issue here was. You didn't link to the test page anywhere. The problem here is that some of the SVG images on the page http://a.deveria.com/svg/css_svg_test.html fail to load with an XML parsing error "Document is empty." I believe I've seen errors like these before, but I'm not sure if there is a bug open for it.

I think webkit is doing ok? but my build certainly not so I need a clue which part that failing/messed (during build process). Initially I think it's the whole svg part, later it seems my webkit still able to render "svg as image" or this is different subject?

Anyway I give one more example from: http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/basic-filters-blend-01-b.html (actually all? tests failed)

Now the svg file did downloaded (if I do reload the svg part and reload percentage showed) but somehow not displayed. This time no GTK error. I really curious what could be wrong with my build.

[Bakhtiar Hasmanan]

Created attachment 78760 [details]
w3c svg filter test

[Bakhtiar Hasmanan]

Created attachment 79236 [details]
Webinspector error and warning screenshot

This error from opening http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/basic-color-prop-01-b.html

XML self-closing tag syntax used on <object>. The tag will not be closed. -> basic-color-prop-01-b.html:58
Resource interpreted as document but transferred with MIME type image/svg+xml. -> color-prop-01-b.svg

[Sergio Villar Senin]

I cannot see any error using r76777

[Bakhtiar Hasmanan]

OK maybe this'd be better closed. My build certainly has an issue, is there a test set that confirm svg feature? I've pass all supplied tests but none seems deal with svg rendering.

[Martin Robinson]

(In reply to comment #8)
> I cannot see any error using r76777

I'm still seeing this with the r76801. Perhaps the issue is the version of libxml installed. I have 2.7.7.dfsg-4ubuntu0.1.

[Sergio Villar Senin]

(In reply to comment #10)
> (In reply to comment #8)
> > I cannot see any error using r76777
> 
> I'm still seeing this with the r76801. Perhaps the issue is the version of libxml installed. I have 2.7.7.dfsg-4ubuntu0.1.

I'm using 2.7.7 too...

[Sergio Villar Senin]

(In reply to comment #11)
> (In reply to comment #10)
> > (In reply to comment #8)
> > > I cannot see any error using r76777
> > 
> > I'm still seeing this with the r76801. Perhaps the issue is the version of libxml installed. I have 2.7.7.dfsg-4ubuntu0.1.
> 
> I'm using 2.7.7 too...

OK, now I see the issue. It's a problem with libsoup that is reporting a problem decoding a data: url used to transfer one of the images. I'll take a look...

[Sergio Villar Senin]

This bug is now tracked in the gnome bugzilla as it's a libsoup bug:
https://bugzilla.gnome.org/show_bug.cgi?id=646896

[Sergio Villar Senin]

(In reply to comment #13)
> This bug is now tracked in the gnome bugzilla as it's a libsoup bug:
> https://bugzilla.gnome.org/show_bug.cgi?id=646896

BTW I was talking about the issue with http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/basic-color-prop-01-b.html.

Regarding http://a.deveria.com/svg/css_svg_test.html it fails also in Chrome and Safari, it's a WebKit issue. There are some related issues
* https://bugs.webkit.org/show_bug.cgi?id=54325
* https://bugs.webkit.org/show_bug.cgi?id=32307
* https://bugs.webkit.org/show_bug.cgi?id=15453

[Bakhtiar Hasmanan]

Based on https://bugs.webkit.org/show_bug.cgi?id=54366 Now I can display http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/basic-color-prop-01-b.html (see screenshot) with webkitgtk-1.2.7 (libsoup-2.32.2). Is it still an issue?

As for http://a.deveria.com/svg/css_svg_test.html you may close this bug then...

[Bakhtiar Hasmanan]

Created attachment 88410 [details]
w3c svg filter test (new)

[Sergio Villar Senin]

(In reply to comment #15)
> Based on https://bugs.webkit.org/show_bug.cgi?id=54366 Now I can display http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/basic-color-prop-01-b.html (see screenshot) with webkitgtk-1.2.7 (libsoup-2.32.2). Is it still an issue?

Looks like it's worning fine, isn't it?