OSX - Compiling training tools for Tesseract 4.0 - pango libraries not found - macos

Question

I am having trouble getting the configure script to recognise the pango libraries i have installed via Homebrew, which are dependancies for the training tools. So far, I've followed all steps detailed here:
https://github.com/tesseract-ocr/tesseract/issues/1453
and here:
https://github.com/tesseract-ocr/tesseract/wiki/Compiling#macos
and after running the configure script, I keep on getting these warnings in the output:
checking for pango >= 1.22.0... no
configure: WARNING: pango 1.22.0 or higher is required, but was not
found.
configure: WARNING: Training tools WILL NOT be built.
configure: WARNING: Try to install libpango1.0-dev package.
checking for cairo... no
configure: WARNING: Training tools WILL NOT be built because of
missing cairo library.
configure: WARNING: Try to install libcairo-dev?? package.
The libraries are located where they should be, at /usr/local/Cellar/pango/1.42.4_1
I've tried running the configure script as per the instructions above:
./configure CPPFLAGS=-I/usr/local/opt/icu4c/include LDFLAGS=-L/usr/local/opt/icu4c/lib
as well as trying to add the path to the linking flags:
./configure LDFLAGS="-L/usr/local/opt/icu4c/lib -L/usr/local/Cellar/pango/1.42.4_1/lib" CPPFLAGS="-I/usr/local/opt/icu4c/include"
and I still can't get the script to see the libraries.
Thanks in advance for any advice on how to solve this.

Answer 1

It looks like pango and/or cairo and maybe other libraries depend on libffi. libffi is called to be keg-only in Homebrew. Even if libffi is installed it is not linked via symbolic-links. When pkg-config tries to find a library, it also checks for its dependencies. Since libffi cannot be located pkg-config tells that cairo is not available. Therefore, one needs to call
brew link libffi --force
This should solve the problem without touching anything extra. At least it worked for my case.

Answer 2

This is a modified version of the answer provided by Boris [Boris: just say the word and I will delete this].
Tested on macOS 10.13.6 High Sierra.
None of the pre-existing answers worked for me - I had to explicitly add libpng to my brew requirements as given by Boris:
brew install cairo pango icu4c autoconf libffi libarchive libpng
export PKG_CONFIG_PATH=\
$(brew --prefix)/lib/pkgconfig:\
$(brew --prefix)/opt/libarchive/lib/pkgconfig:\
$(brew --prefix)/opt/icu4c/lib/pkgconfig:\
$(brew --prefix)/opt/libffi/lib/pkgconfig:\
$(brew --prefix)/opt/libpng/lib/pkgconfig
./configure
Note: If you already have some of the above you might rather want to run brew reinstall in place of brew install, at the risk of damaging other projects.
I didn't need to supply any flags to configure.
The aim is of course to get the training tools to compile so that combine_tessdata, wordlist2dawg, dawg2wordlist etc. are available (I haven't found macOS binaries for these anywhere).
The full instructions then become:
git clone git#github.com:tesseract-ocr/tesseract.git
git checkout tags/4.1.0
cd tesseract
./autogen.sh
brew install cairo pango icu4c autoconf libffi libarchive libpng
export PKG_CONFIG_PATH=\
$(brew --prefix)/lib/pkgconfig:\
$(brew --prefix)/opt/libarchive/lib/pkgconfig:\
$(brew --prefix)/opt/icu4c/lib/pkgconfig:\
$(brew --prefix)/opt/libffi/lib/pkgconfig:\
$(brew --prefix)/opt/libpng/lib/pkgconfig
./configure
make -j
sudo make install
# -> You should now have access to tesseract
make training
sudo make training-install
# -> You should now have access to the training tools: combine_tessdata, wordlist2dawg, dawg2wordlist, etc.
# If not, check /usr/local/bin/ is on your PATH, i.e.
export PATH=/usr/local/bin/:$PATH
Footnote: How I diagnosed the libpng problem:
As stated above and here, the error thrown is not very helpful:
make training
Need to reconfigure project, so there are no errors
Digging a little deeper, configure throws some warnings along the lines of:
configure: WARNING: pango 1.22.0 or higher is required, but was not found.
configure: WARNING: Training tools WILL NOT be built.
configure: WARNING: Try to install libpango1.0-dev package.
checking for cairo... no
configure: WARNING: Training tools WILL NOT be built because of missing cairo library.
configure: WARNING: Try to install libcairo-dev?? package.
checking that generated files are newer than configure... done
I also had one for cairo
So you have to run configure in debug mode:
./configure --enable-debug
And check config.log by greping for pango (say):
$PKG_CONFIG --exists --print-errors "pango >= 1.22.0"
That is the line that throws the warning when configure is executed.
You can now execute that last line manually:
pkg-config --exists --print-errors pango
dyld: Symbol not found: __cg_png_create_info_struct
Referenced from: /System/Library/Frameworks/ImageIO.framework/Versions/A/ImageIO
Expected in: /usr/local/lib/libPng.dylib
in /System/Library/Frameworks/ImageIO.framework/Versions/A/ImageIO
Abort trap: 6
which is how I traced the libpng cause. (In my case I had to reinstall/relink libpng because of various long-running clashes associated with gnuplot/octave/aquaterm etc. etc.).
HTH someone

Answer 3

This is the most elegant solution to the problem I found:
brew install cairo pango icu4c autoconf libffi libarchive libpng
export PKG_CONFIG_PATH=\
$(brew --prefix)/lib/pkgconfig:\
$(brew --prefix)/opt/libarchive/lib/pkgconfig:\
$(brew --prefix)/opt/icu4c/lib/pkgconfig:\
$(brew --prefix)/opt/libffi/lib/pkgconfig:\
$(brew --prefix)/opt/libpng/lib/pkgconfig
./configure
https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#on-macos-mojave-with-homebrew

Answer 4

I fixed this by copying the pkgconfig files under libffi and icu4c
(find the path by typing brew link libffi icu4c in your terminal)
and pasting them at /usr/local/lib/pkgconfig and then set
PKG_CONFIG_PATH=/usr/local/lib/pkgconfig