Tesseract 4.0 on Alpine Linux Docker Container

[Kalven Schraut]

I am attempting to use tesseract's API in my project and everything works as expected on ubuntu when running the code, but I am receiving a seg fault when I moved everything over to an alpine docker container.

The backtrace from the segfault:

#0  0x00007ffff2c4a50a in ?? () from /usr/lib/libgomp.so.1

#1  0x00007ffff2c45d02 in GOMP_parallel () from /usr/lib/libgomp.so.1

#2  0x00007ffff492cfea in tesseract::FullyConnected::Forward (this=0x5555577dbb20, debug=<optimized out>, input=..., input_transpose=<optimized out>,

    scratch=0x5555577f9fc8, output=0x55555877b6a0) at fullyconnected.cpp:140

#3  0x00007ffff49598ff in tesseract::Series::Forward (this=0x555557803f60, debug=<optimized out>, input=..., input_transpose=<optimized out>,

    scratch=0x5555577f9fc8, output=0x55555877b6a0) at series.cpp:123

#4  0x00007ffff49598ff in tesseract::Series::Forward (this=0x555557803d60, debug=<optimized out>, input=..., input_transpose=<optimized out>,

    scratch=0x5555577f9fc8, output=0x7fffffffc380) at series.cpp:123

#5  0x00007ffff493b8ce in tesseract::LSTMRecognizer::RecognizeLine (this=this@entry=0x5555577f9c80, image_data=..., invert=invert@entry=true,

    debug=debug@entry=false, re_invert=re_invert@entry=false, upside_down=upside_down@entry=false, scale_factor=0x7fffffffc35c, inputs=0x7fffffffc410,

    outputs=0x7fffffffc380) at lstmrecognizer.cpp:256

#6  0x00007ffff493c4d0 in tesseract::LSTMRecognizer::RecognizeLine (this=0x5555577f9c80, image_data=..., invert=invert@entry=true, debug=false,

    worst_dict_cert=worst_dict_cert@entry=-3.5714285373687744, line_box=..., words=words@entry=0x7fffffffc600) at lstmrecognizer.cpp:190

#7  0x00007ffff480978f in tesseract::Tesseract::LSTMRecognizeWord (this=this@entry=0x555557782420, block=..., row=row@entry=0x55555861c2e0, word=<optimized out>,

    words=words@entry=0x7fffffffc600) at linerec.cpp:241

#8  0x00007ffff47ef729 in tesseract::Tesseract::classify_word_pass1 (this=0x555557782420, word_data=..., in_word=0x5555585bd4e0, out_words=0x7fffffffc600)

    at control.cpp:1373

#9  0x00007ffff47f09a5 in tesseract::Tesseract::RetryWithLanguage (this=0x555557782420, word_data=..., recognizer=<optimized out>, debug=debug@entry=false,

    in_word=0x5555585bd4e0, best_words=0x7fffffffc6e0) at control.cpp:898

#10 0x00007ffff47f113b in tesseract::Tesseract::classify_word_and_language (this=this@entry=0x555557782420, pass_n=pass_n@entry=1,

    pr_it=pr_it@entry=0x7fffffffc850, word_data=word_data@entry=0x55555863ac08) at control.cpp:1314

#11 0x00007ffff47f463c in tesseract::Tesseract::RecogAllWordsPassN (this=this@entry=0x555557782420, pass_n=pass_n@entry=1, monitor=monitor@entry=0x0,

    pr_it=pr_it@entry=0x7fffffffc850, words=words@entry=0x7fffffffc830) at control.cpp:265

#12 0x00007ffff47f612d in tesseract::Tesseract::recog_all_words (this=0x555557782420, page_res=0x5555585d6160, monitor=monitor@entry=0x0,

    target_word_box=target_word_box@entry=0x0, word_config=word_config@entry=0x0, dopasses=dopasses@entry=0) at control.cpp:352

I first tried installing the tesseract-git package in alpine where I noticed the issue so I just finished compiling the master branch of tesseract-ocr and I am still receiving the seg fault.

Also my compiled version of tesseract runs fine through the CLI.

I am lost as to what else could be the problem and would appreciate any help/direction on how to solve this issue.

[Kalven Schraut]

Looks like the issue is from

#pragma omp parallel for num_threads(kNumThreads) at

https://github.com/tesseract-ocr/tesseract/blob/master/lstm/fullyconnected.cpp#L140

anyone more familiar with openMP know a possible reason for a seg fault there?

[Kalven Schraut]

also looks like calling configure with the --disable-openmp prevents the seg fault which allows me to continue to use tesseract for now would still like to figure out why the seg fault is occuring with multiple threads.