Failing unit test in Mac OS X Yosemite 10.10.1

[Stephan Bergmann]

On 12/13/2014 12:44 AM, Renato Ferreira wrote:
> commit b1b29ae795ff86097b21907f814d94bd5488929d
> Author: Stephan Bergmann <sbergman at redhat.com <mailto:sbergman at redhat.com>>
> Date:   Thu Nov 20 12:06:11 2014 +0100
>
>      ...at least, that's the value for me on Mac with retina
>      Change-Id: I041b497bab9415b2b33d6b4b91f3c58ea9dbc05f
>
> diff --git a/sw/qa/extras/ooxmlexport/ooxmlexport.cxx
> b/sw/qa/extras/ooxmlexport/ooxmlexport.cxx
> index 519ad65..96b0d0ee 100644
> --- a/sw/qa/extras/ooxmlexport/ooxmlexport.cxx
> +++ b/sw/qa/extras/ooxmlexport/ooxmlexport.cxx
> @@ -484,9 +484,7 @@ DECLARE_OOXMLEXPORT_TEST(testCropPixel,
> "crop-pixel.docx")
>           // This is 17667 in the original document, was 504666 (so the
> image
>           // become invisible), now is around 19072.
>           (void) pXmlDoc;
> -#ifndef MACOSX
> -        CPPUNIT_ASSERT(getXPath(pXmlDoc, "//a:srcRect", "l").toInt32()
> < 20000);
> -#endif
> +        CPPUNIT_ASSERT(getXPath(pXmlDoc, "//a:srcRect", "l").toInt32()
> <= 21856);
>       }
>   }
>
> My Mac (Retina 13") gives the value 22452, but I'm not able to tell
> which value is "valid", or on which Mac with retina the other value was
> produced. Does anybody else use OS X Yosemite here?

21856 is what I consistently get on a 15" retina Mac, building against 
10.10 SDK.

And as 21856 appears be be reasonably close still to 17667, at least in 
comparison to 504666, I thought I give that commit a try as an 
improvement over the blunt 
<http://cgit.freedesktop.org/libreoffice/core/commit/?id=f5296655650e75fee259acf1959c15eb2cea3a1e> 
"disable specific oox_export unit for macosx," without further insight 
into why values actually differ though.