Segfault with new image and rus+mrz in recent main

[marcreichman-pfi]

Current Behavior

With the attached image, and recent main (9f17a3fd) I receive a segfault in Release (SIGILL in Debug) with the mrz and rus langages. Both are fast. Rus is official and mrz is available here

Release:

<snip>/tesseract-fork/cmake-build-release/bin/tesseract <snip>/62553353F7495226759F9644269AA485-sample_000101.jpg - --tessdata-dir <snip>/tessdata/ -l mrz+rus
Estimating resolution as 971

Process finished with exit code 139 (interrupted by signal 11:SIGSEGV)

Debug:

Estimating resolution as 971
ELIST2_ITERATOR::add_before_stay_put:Error:Attempting to add an element with non nullptr links, to a list

Process finished with exit code 132 (interrupted by signal 4:SIGILL)

Expected Behavior

No segfault.

Suggested Fix

No response

tesseract -v

tesseract 5.5.0-26-g9f17a
 leptonica-1.82.0
  libgif 5.1.9 : libjpeg 8d (libjpeg-turbo 2.1.1) : libpng 1.6.37 : libtiff 4.3.0 : zlib 1.2.11 : libwebp 1.2.2 : libopenjp2 2.4.0
 Found AVX
 Found SSE4.1
 Found OpenMP 201511

Operating System

Ubuntu 22.04 Jammy

Other Operating System

This is running via WSL but it does not appear related to the issue.

uname -a

Linux hostname 5.10.16.3-microsoft-standard-WSL2 #1 SMP Fri Apr 2 22:23:49 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

Compiler

GCC 11.4

CPU

Intel(R) Core(TM) i7-3720QM CPU @ 2.60GHz

Virtualization / Containers

No explicit VM except for WSL2

Other Information

This works in 5.5.0 tag

[egorpugin]

Can you run your command under gdb and post stack trace? (debug tess binary)

gdb tesseract
set args *your args to tesseract*
r

* wait for the crash *

bt

[marcreichman-pfi]

@egorpugin This is with a debug build, hence the SIGILL instead of SIGSEGV:

(gdb) set args ~/dev/testimages/new-segfault-20241125.jpg - --tessdata-dir <snip>/tessdata/ -l mrz+rus
(gdb) r
Starting program: /root/dev/tesseract/build-debug/bin/tesseract ~/dev/testimages/new-segfault-20241125.jpg - --tessdata-dir <snip>/tessdata/ -l mrz+rus
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Estimating resolution as 971
[New Thread 0x7ffff73c6640 (LWP 1893)]
[New Thread 0x7ffff6bc5640 (LWP 1894)]
[New Thread 0x7ffff63c4640 (LWP 1895)]
ELIST2_ITERATOR::add_before_stay_put:Error:Attempting to add an element with non nullptr links, to a list

Thread 1 "tesseract" received signal SIGILL, Illegal instruction.
tesseract::ERRCODE::error (this=this@entry=0x5555558a1308 <tesseract::STILL_LINKED>, caller=caller@entry=0x5555557f8f00 "ELIST2_ITERATOR::add_before_stay_put", action=action@entry=tesseract::ABORT, format=format@entry=0x0) at /root/dev/tesseract/src/ccutil/errcode.cpp:78
78            __builtin_trap();
(gdb) bt
#0  tesseract::ERRCODE::error (this=this@entry=0x5555558a1308 <tesseract::STILL_LINKED>, caller=caller@entry=0x5555557f8f00 "ELIST2_ITERATOR::add_before_stay_put",
    action=action@entry=tesseract::ABORT, format=format@entry=0x0) at /root/dev/tesseract/src/ccutil/errcode.cpp:78
#1  0x000055555558a98d in tesseract::ERRCODE::error (action=tesseract::ABORT, caller=0x5555557f8f00 "ELIST2_ITERATOR::add_before_stay_put", this=0x5555558a1308 <tesseract::STILL_LINKED>)
    at /root/dev/tesseract/src/ccutil/errcode.h:47
#2  tesseract::IntrusiveList<tesseract::WERD>::Iterator::add_before_stay_put (new_element=0x555556132100, this=0x7fffffffd8f0) at /root/dev/tesseract/src/ccutil/elst2.h:387
#3  tesseract::PAGE_RES_IT::ReplaceCurrentWord (this=this@entry=0x7fffffffdc10, words=words@entry=0x7fffffffdaa0) at /root/dev/tesseract/src/ccstruct/pageres.cpp:1469
#4  0x0000555555612f42 in tesseract::Tesseract::classify_word_and_language (this=this@entry=0x7ffff73c7010, pass_n=pass_n@entry=1, pr_it=pr_it@entry=0x7fffffffdc10,
    word_data=word_data@entry=0x555556ba5cd0) at /root/dev/tesseract/src/ccmain/control.cpp:1367
#5  0x000055555561a01f in tesseract::Tesseract::RecogAllWordsPassN (this=this@entry=0x7ffff73c7010, pass_n=pass_n@entry=1, monitor=monitor@entry=0x0, pr_it=pr_it@entry=0x7fffffffdc10,
    words=words@entry=0x7fffffffdbf0) at /root/dev/tesseract/src/ccmain/control.cpp:255
#6  0x000055555561a456 in tesseract::Tesseract::recog_all_words (this=0x7ffff73c7010, page_res=0x555556bb7160, monitor=monitor@entry=0x0, target_word_box=target_word_box@entry=0x0,
    word_config=word_config@entry=0x0, dopasses=dopasses@entry=0) at /root/dev/tesseract/src/ccmain/control.cpp:345
#7  0x00005555555d5553 in tesseract::TessBaseAPI::Recognize (this=this@entry=0x7fffffffe2e0, monitor=monitor@entry=0x0) at /root/dev/tesseract/src/api/baseapi.cpp:833
#8  0x00005555555d57e3 in tesseract::TessBaseAPI::ProcessPage (this=this@entry=0x7fffffffe2e0, pix=0x5555558d6020, page_index=page_index@entry=0,
    filename=filename@entry=0x7fffffffe783 "/root/dev/testimages/new-segfault-20241125.jpg", retry_config=retry_config@entry=0x0, timeout_millisec=timeout_millisec@entry=0,
    renderer=0x5555558d2740) at /root/dev/tesseract/src/api/baseapi.cpp:1218
#9  0x00005555555d68e4 in tesseract::TessBaseAPI::ProcessPagesInternal (this=this@entry=0x7fffffffe2e0, filename=0x7fffffffe783 "/root/dev/testimages/new-segfault-20241125.jpg",
    retry_config=retry_config@entry=0x0, timeout_millisec=timeout_millisec@entry=0, renderer=0x5555558d2740) at /root/dev/tesseract/src/api/baseapi.cpp:1181
#10 0x00005555555d69ea in tesseract::TessBaseAPI::ProcessPages (this=this@entry=0x7fffffffe2e0, filename=<optimized out>, retry_config=retry_config@entry=0x0,
    timeout_millisec=timeout_millisec@entry=0, renderer=<optimized out>) at /root/dev/tesseract/src/api/baseapi.cpp:998
#11 0x000055555556d6c3 in main (argc=<optimized out>, argv=<optimized out>) at /usr/include/c++/11/bits/unique_ptr.h:173
(gdb)

[egorpugin]

Now please download this file
https://github.com/tesseract-ocr/tesseract/blob/5c78037bd3d79dbb71bdccf6333a8d94b530fd19/src/ccutil/helpers.h
replace src/ccutil/helpers.h with it
rebuild debug build and try again.

[marcreichman-pfi]

@egorpugin This one works fine now.

On a hunch I ran the images from #4148, and those now fail again.

I am sure from your comments you understand, the random stuff is masking the various failures that may actually be happening under the hood.

[egorpugin]

Yes, that is what I'm testing.

[egorpugin]

It is better to keep that bunch of tickets open until the real issue(s) are fixed.

[amitdo]

Try to add a black frame to the image.

convert in.png -bordercolor Black -border 10x10 in-black-bprder.png

[marcreichman-pfi]

Hi @amitdo, I am aware that subtle changes to the images usually make things work. We've had a series of images while using high throughput through many videos and other images. The segfaults and aborts are popping up from time to time which is why we're filing the tickets.

Thanks!

[amitdo]

Please try it and tell me if after adding the black frame you still get a segfault.

[marcreichman-pfi]

Hi @amitdo

With the border it did not create a segfault. I also tried some others.

convert in.jpg -bordercolor Black -border 10x10 in-black-border.jpg <- no crash
convert in.jpg -bordercolor Black -border 10x10 in-black-border.png <- no crash
convert in.jpg in-copy.jpg <- no crash
convert in.jpg in-copy.png <- crash

All of this makes sense based on things I've seen testing all these tickets.

[amitdo]

Like the issues in boxClipToRectangle , Tesseract does not like images where pixels from the foreground touch the image frame.