centos 7.9 tesseract5.2 cmake failed! boxchar.cpp:65:75: error...

[wuyang-dl]

error INFO:
/home/engine/wy/3rd/tesseract-5.2.0/src/training/pango/boxchar.cpp: In member function ‘void tesseract::BoxChar::GetDirection(int*, int*) const’:
/home/engine/wy/3rd/tesseract-5.2.0/src/training/pango/boxchar.cpp:65:75: error: ‘U_RIGHT_TO_LEFT_ISOLATE’ was not declared in this scope; did you mean ‘U_RIGHT_TO_LEFT_OVERRIDE’?
65 | if (dir == U_RIGHT_TO_LEFT || dir == U_RIGHT_TO_LEFT_ARABIC || dir == U_RIGHT_TO_LEFT_ISOLATE) {
| ^~~~~~~~~~~~~~~~~~~~~~~
| U_RIGHT_TO_LEFT_OVERRIDE

Environment

• Tesseract Version: 5.2.0
• Commit Number: tag 5.2.0
• Platform:centos7.9

error.log shows details
error.log

[zdenop]

Did you search the issue tracker before posting the issue? .e.g. #1374

[amitdo]

The CMake build should do what the Autotools build is doing: check for icu >=52.1 and refuse to build the training tools if this requirement is not met.

[zdenop]

@amitdo : it does: https://github.com/tesseract-ocr/tesseract/runs/7640401854?check_suite_focus=true#step:6:82

[zdenop]

here is relevant part of check:

tesseract/src/training/CMakeLists.txt

Lines 71 to 77 in 94b9ca4

	if(NOT SW_BUILD)
	if(PKG_CONFIG_FOUND)
	pkg_check_modules(ICU REQUIRED IMPORTED_TARGET icu-uc icu-i18n)
	else()
	find_package(ICU 52.1 COMPONENTS uc i18n)
	endif()
	endif()

In provided log there is no info about ICU checks...

[amitdo]

Thanks, Zdenko. I only looked at the MakeLists.txt located in the root directory. I forgot that there is another one in the training dir.

[amitdo]

if(PKG_CONFIG_FOUND) 
     pkg_check_modules(ICU REQUIRED IMPORTED_TARGET icu-uc icu-i18n)

There is no version check here.

[zdenop]

@amitdo : you miss the point: according reporter log there was not check for ICU. So putting there any version does not solve reporter problem. I wander how reporter managed to skip ICU presence. Plus ICU 52.1 was released 2013-10-09, so I really wonder if somebody is using older version than that...

[amitdo]

Plus ICU 52.1 was released 2013-10-09, so I really wonder if somebody is using older version than that...

http://mirror.centos.org/centos/7.9.2009/os/x86_64/Packages/

libicu-50.2-4.el7_7.i686.rpm

[zdenop]

So the solution is to use recent and well-maintained OS/distribution.

[wuyang-dl]

So the solution is to use recent and well-maintained OS/distribution.

hi, I use libicui18n.so( with yum install libicu-devel cmd), for tess check script shows error info:
-- Checking for modules 'icu-uc;icu-i18n'
-- No package 'icu-uc' found
-- No package 'icu-i18n' found

after installing libicu-devel, pango-devel and cairo-devel,
cmake -D CMAKE_INSTALL_PREFIX=/usr/local -D CMAKE_BUILD_TYPE=RELEASE -D BUILD_SHARED_LIBS=ON ..

then error occurred(Linux version 3.10.0-1160.71.1.el7.x86_64 ([email protected]) (gcc version 4.8.5 20150623 (Red Hat 4.8.5-44) (GCC) ) #1 SMP Tue Jun 28 15:37:28 UTC 2022).

I tried ubuntu(Linux version 5.4.0-122-generic (buildd@lcy02-amd64-035) (gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu118.04)) #13818.04.1-Ubuntu SMP Fri Jun 24 14:14:03 UTC 2022), tess is ok

tks

[amitdo]

You can try to build with autotools instead of CMake.

GCC 4.8 is not supported in Tesseract 5.x.

RHEL/Centos have newer GCC versions in their repos:
http://mirror.centos.org/centos/7.9.2009/sclo/x86_64/rh/Packages/d/

So you can install GCC 11.

[amitdo]

Do you plan to train your own model using Tesseract?

If not, ICU, Pango and Cairo are not required.

Autotools: Pango, Cairo and ICU only required by training tools

I don't know if CMake behave in the same way.

[zdenop]

CMake behaves in the same way as Autotools. From the above information, I am sure that the reporter must modify CMake files to avoid these checks (ICU and GCC). So this is tesseract problem, but a user problem (try to compile tesseract on too with too old software)

[amitdo]

CentOS Linux 7 reached EOL on June 30, 2024.