Details
-
Bug
-
Resolution: Done
-
P2: Important
-
5.0.0
-
None
-
af6c1a32982564035f3f9d1c0887c9d3fc0e5bbb
Description
As of 30eff894d9b31e0c800901e7334fed2401be0f1c in qtbase, tst_qtextscriptengine says in a comment:
// This test depends on the fonts in the following package being installed: // http://people.freedesktop.org/~hausmann/harfbuzz-test-fonts-0.1.tar.bz2
This seems to be a broken link.
Also, the test currently will run if any font on the system appears to provide the tested scripts, but comments on http://qt.gitorious.org/qt/qt/merge_requests/2484 imply that only specific versions of some fonts are expected to let the test pass. Since the test doesn't verify these specific versions of the fonts, the test may fail when run on a system with other fonts which happen to provide the scripts.
This currently happens in our CI, on Ubuntu 10.04
QDEBUG : tst_QTextScriptEngine::gurmukhi() Lohit Punjabi: shaping of string a15 a4d a39 failed, nglyphs=2, expected 2 QDEBUG : tst_QTextScriptEngine::gurmukhi() glyph result = 3f 91 FAIL! : tst_QTextScriptEngine::gurmukhi() 'shaping(f, s)' returned FALSE. () Loc: [tst_qtextscriptengine.cpp(562)]
The above could either be a genuine bug, or it could be that "Lohit Punjabi" is provided by a font whose implementation details doesn't match exactly what the test is coded for.