Enable pango_font_info_test and fix crash for MacOS with M1

[stweil]

No description provided.

[stweil]

pango_font_info_test passes on Linux. On MacOS with M1 it at least no longer crashes and passes most parts of that test.

[stweil]

See also issue #2599.

[zdenop]

build is failing for ubuntu:

[ 92%] Building CXX object src/training/CMakeFiles/text2image.dir/pango_font_info.cpp.o
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp: In function ‘PangoGlyph tesseract::get_glyph(PangoFont*, gunichar)’:
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:76:3: error: ‘hb_font_t’ was not declared in this scope
   hb_font_t* hb_font = pango_font_get_hb_font(font);
   ^
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:76:14: error: ‘hb_font’ was not declared in this scope
   hb_font_t* hb_font = pango_font_get_hb_font(font);
              ^
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:76:51: error: ‘pango_font_get_hb_font’ was not declared in this scope
   hb_font_t* hb_font = pango_font_get_hb_font(font);
                                                   ^
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:77:3: error: ‘hb_codepoint_t’ was not declared in this scope
   hb_codepoint_t glyph;
   ^
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:78:43: error: ‘glyph’ was not declared in this scope
   hb_font_get_nominal_glyph(hb_font, wc, &glyph);
                                           ^
/home/travis/build/tesseract-ocr/tesseract/src/training/pango_font_info.cpp:78:48:m error: ‘hb_font_get_nominal_glyph’ was not declared in this scope
   hb_font_get_nominal_glyph(hb_font, wc, &glyph);

[stweil]

The new code requires Pango 1.44 for pango_font_get_hb_font. I now added the old code as a fallback for older Pango versions, so please retry.

[amitdo]

pango_font_info_test passes on Linux. On MacOS with M1 it at least no longer crashes and passes most parts of that test.

Which parts do not pass? Maybe you can add TODO above each one?

[stweil]

It looks like on MacOS the test sees all MacOS fonts instead of the limited set which is installed in test/testing. I don't know the reason for that, but it explains the failure of several test steps.

[  PASSED  ] 8 tests.
[  FAILED  ] 6 tests, listed below:
[  FAILED  ] PangoFontInfoTest.CanRenderString
[  FAILED  ] PangoFontInfoTest.CanRenderLigature
[  FAILED  ] FontUtilsTest.DoesFindAvailableFonts
[  FAILED  ] FontUtilsTest.DoesDetectMissingFonts
[  FAILED  ] FontUtilsTest.DoesListAvailableFonts
[  FAILED  ] FontUtilsTest.DoesFailToSelectFont

[amitdo]

The text renderer was designed to work with the FontConfig backend, even on macOS, and it appears that your pango version uses the CoreText backend.

[stweil]

It's from Homebrew, and that should work with FontConfig.
text2image --list_available_fonts --fonts_dir tesseract/test/testing works.

[stweil]

[...] it appears that your pango version uses the CoreText backend.

You are right, thank you. The test passes with export PANGOCAIRO_BACKEND=fc.

[amitdo]

It works with text2image thanks to this code:

tesseract/src/training/text2image.cpp

Lines 720 to 726 in 45413e6

	int main(int argc, char** argv) {
	// Respect environment variable. could be:
	// fc (fontconfig), win32, and coretext
	// If not set force fontconfig for Mac OS.
	// See https://github.com/tesseract-ocr/tesseract/issues/736
	char* backend;
	backend = getenv("PANGOCAIRO_BACKEND");

[zdenop]

thanks