text2image needs pango_training.so which in turn is not installed

[leleliu008]

Current Behavior

text2image needs pango_training.so which in turn is not installed

Expected Behavior

pango_training.so installed.

Suggested Fix

https://github.com/tesseract-ocr/tesseract/blob/main/src/training/CMakeLists.txt#L405-L412

TARGETS text2image

change to:

TARGETS text2image pango_training

tesseract -v

No response

Operating System

Ubuntu 22.04 Jammy

Other Operating System

Android

uname -a

Linux leleliu008 6.5.0-28-generic #29-Ubuntu SMP PREEMPT_DYNAMIC Thu Mar 28 23:46:48 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

Compiler

Android NDK r26d (clang version 17.0.2)

CPU

aarch64

Virtualization / Containers

No response

Other Information

source: https://github.com/tesseract-ocr/tesseract/archive/refs/tags/5.3.4.tar.gz

[stweil]

Did you build text2image yourself (if yes: how)? And how did you install it?

[leleliu008]

@stweil

Yes, I build it for Android.

I din't give you the reproduce steps, because it is very clear if you look at the source code https://github.com/tesseract-ocr/tesseract/blob/main/src/training/CMakeLists.txt#L405-L412

My build instuctions:

cmake \
        -Wno-dev \
        -DBUILD_TESTING=OFF \
        -DCMAKE_INSTALL_LIBDIR='/github/home/.ndk-pkg/installed/android-21-arm64-v8a/e485a3cff7749b1b6d89b185304b1ff469c336fb646f1b164b65578e9bad7ccd/lib' \
        -DCMAKE_INSTALL_PREFIX='/github/home/.ndk-pkg/installed/android-21-arm64-v8a/e485a3cff7749b1b6d89b185304b1ff469c336fb646f1b164b65578e9bad7ccd' \
        -DCMAKE_TOOLCHAIN_FILE='/github/home/.ndk-pkg/run/750/android-21-arm64-v8a/tesseract/android.toolchain.cmake' \
        -DCMAKE_VERBOSE_MAKEFILE='ON' \
        -DCMAKE_COLOR_MAKEFILE='ON' \
        -S /github/home/.ndk-pkg/run/750/android-21-arm64-v8a/tesseract/src \
        -B /github/home/.ndk-pkg/run/750/android-21-arm64-v8a/tesseract/src/_ \
        -DENABLE_LTO=OFF \
        -DUSE_SYSTEM_ICU=ON \
        -DDISABLE_ARCHIVE=OFF \
        -DDISABLE_CURL=OFF \
        -DBUILD_TESTS=OFF \
        -DBUILD_TRAINING_TOOLS=ON \
        -DOPENMP_BUILD=ON \
        -DLEPT_TIFF_RESULT=0

[stweil]

I am curious:

Are there useful applications for training on Android?
And how did you create that nice dependency diagram?

[leleliu008]

Are there useful applications for training on Android?

I found this when I built it for Android, I think this problem is nothing to do with specific platform.

And how did you create that nice dependency diagram?

I create it in dot https://graphviz.org/doc/info/command.html

[zdenop]

I found this when I built it for Android, I think this problem has nothing to do with a specific platform.

Why do you try to build training tools if you are not able/want to use them on your platform? Just to get extra issues?

I din't give you the reproduce steps, because it is very clear if you look at the source code https://github.com/tesseract-ocr/tesseract/blob/main/src/training/CMakeLists.txt#L405-L412

I see no reason why to install pango_training:

$ ldd build/bin/text2image | grep -i pango
        libpangocairo-1.0.so.0 => /lib/arm-linux-gnueabihf/libpangocairo-1.0.so.0 (0xb6660000)
        libpango-1.0.so.0 => /lib/arm-linux-gnueabihf/libpango-1.0.so.0 (0xb65f0000)
        libpangoft2-1.0.so.0 => /lib/arm-linux-gnueabihf/libpangoft2-1.0.so.0 (0xb3980000)

The reason is very simple - the official build process for training tools uses static linking.

[leleliu008]

If you want to static linking, you should change add_library(pango_training ${pango_training_src}) to add_library(pango_training STATIC ${pango_training_src})

[leleliu008]

It a feature not a bug, so I'm closing.