Stephan Bergmann 9e83d35c4a Make CppunitTest_svgio CPPUNIT_TEST_NAME=Test::testCssClassRedefinition robust ...

That recently introduced test systematically failed on my Fedora 36 machine with

> /home/sbergman/lo2/core/test/source/xmltesttools.cxx:89:(anonymous namespace)::Test::testCssClassRedefinition
> equality assertion failed
> - Expected: 1
> - Actual  : 0
> - In <>, XPath '/primitive2D/transform/textsimpleportion[1]' number of nodes is incorrect

because the dumped data was

> <?xml version="1.0" encoding="UTF-8"?>
> <primitive2D>
>  <transform xy11="26" xy12="0" xy13="0" xy21="0" xy22="26" xy23="0" xy31="0" xy32="0" xy33="1">
>   <mask>
>    <polypolygon height="100" width="50" minx="0" miny="0" maxx="50" maxy="100" path="m0 0h50v100h-50z">
>     <polygon>
>      <point x="0" y="0"/>
>      <point x="50" y="0"/>
>      <point x="50" y="100"/>
>      <point x="0" y="100"/>
>     </polygon>
>    </polypolygon>
>    <textsimpleportion width="16" height="16" x="20" y="20" text="test" fontcolor="#ff0000" familyname="Sans"/>
>   </mask>
>  </transform>
> </primitive2D>

rather than the expected

> <?xml version="1.0" encoding="UTF-8"?>
> <primitive2D>
>  <transform xy11="26" xy12="0" xy13="0" xy21="0" xy22="26" xy23="0" xy31="0" xy32="0" xy33="1">
>   <textsimpleportion width="16" height="16" x="20" y="20" text="test" fontcolor="#ff0000" familyname="Sans"/>
>   <hiddengeometry>
>    <polygonhairline color="#000000">
>     <polygon>0,0 50,0 50,100 0,100</polygon>
>    </polygonhairline>
>   </hiddengeometry>
>  </transform>
> </primitive2D>

because in SvgSvgNode::decomposeSvgNode (svgio/source/svgreader/svgsvgnode.cxx)
aSvgCanvasRange = 50x100@(0,0) vs. aContentRange = 31x11@(20,9) did not match
the

>                         if(aSvgCanvasRange.isInside(aContentRange))

branch ("no clip needed, but an invisible HiddenGeometryPrimitive2D...") but
rather the

>                         else if(aSvgCanvasRange.overlaps(aContentRange))

branch ("Clip is necessary") introducing the additional <mask> element.  The
size of aContentRange (and thus whether it is contained in or merely overlaps
aSvgCanvasRange) is apparently affected by what actual font gets substituted for
"Sans" in any given environment.

The only font that is guaranteed to always be available and is not subject to
--with-fonts (which isn't already demanded via gb_CppunitTest_use_more_fonts in
svgio/CppunitTest_svgio.mk) appears to be "Open Symbol" (cf.
extras/Package_fonts.mk), so use that (even if it doesn't have glyphs for the
given text, but which should be irrelevant for this test).

Change-Id: If6f6fdc481b313aa08803b4b3f406e01ff0be1ac
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/136350
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>

2022-06-23 23:03:21 +02:00

14 lines

497 B

XML

Raw Blame History