Tesseract hangs

[vmato]

Environment

• Tesseract Version:

tesseract 5.2.0
leptonica-1.82.0
libgif 5.2.1 : libjpeg 8d (libjpeg-turbo 2.1.4) : libpng 1.6.37+apng : libtiff 4.4.0 : zlib 1.2.12 : libwebp 1.2.4
Found OpenMP 201811
Found libarchive 3.6.1 zlib/1.2.12 liblzma/5.2.5 bz2lib/1.0.8 liblz4/1.9.4 libzstd/1.5.2
Found libcurl/7.85.0 OpenSSL/1.1.1o zlib/1.2.12 libpsl/0.21.1 (+libidn2/2.3.3) libssh2/1.10.0 nghttp2/1.48.0

• Platform:
  FreeBSD hostname.com 13.1-RELEASE-p3 FreeBSD 13.1-RELEASE-p3 GENERIC amd64

Current Behavior:

Tesseract hangs without any result. Hanged process even can not be killed.
Commands like tesseract anyimage.png stdout or tesseract -l eng+rus anyimage.png stdout

Expected Behavior:

Tesseract works (produces result and exits)

[stweil]

Did you install the Tesseract which is provided by FreeBSD?

Can you run it with a debugger and get a stack trace when it is hanging? Or build it from source and debug the hanging process?

[vmato]

Yes, Tesseract is from binary packages (pkg install tesseract-5.2.0_1)

It's not possible to attach debugger or dump process with kill -QUIT.
Related discussion https://forums.freebsd.org/threads/process-hangs-in-run-state-and-can-not-be-killed-or-debugged.87035/

[amitdo]

Commands like tesseract anyimage.png stdout or tesseract -l eng+rus anyimage.png stdout

Does this hang happen with any image (as the name of the image indicates)?

Does this hang happen with this command:

Tesseract /path/to/myimage.png /path/to/output_filename_without_extension
?

[amitdo]

You can also try to prepend this to the tesseract command:

OMP_THREAD_LIMIT=1

[stweil]

Also interesting: is this issue limited to one specific installation / computer, or does it occur with any installation of FreeBSD / on several or all computers?

[amitdo]

Another thing I just noticed.

According to your tesseract -v output, tesseract does not detect any SIMD instruction. not even SSE4.1. How old is your machine? Do you run this in the cloud? or locally inside a VM? Did you somehow disabled SIMD?

[stweil]

That's indeed very strange because any amd64 platform should at least show support for SSE 4.1.

[amitdo]

any amd64 platform should at least show support for SSE 4.1

First 64-bit CPU from AMD came out in 2003, a few years before SSE4.1 was introduced.

Still, in practice you are right (unless the machine is antique).

https://en.wikipedia.org/wiki/X86-64

https://developers.redhat.com/blog/2021/01/05/building-red-hat-enterprise-linux-9-for-the-x86-64-v2-microarchitecture-level#background_of_the_x86_64_microarchitecture_levels

https://www.phoronix.com/news/GCC-11-x86-64-Feature-Levels

https://www.phoronix.com/news/LLVM-12-Clang-12-Feature-Over

[vmato]

Does this hang happen with any image
We've tried several images. Always with same result

tesseract /home/user/test/test.png /home/user/test/test
Same freeze. But empty output file test.txt was created

OMP_THREAD_LIMIT=1
Didn't get any visible changes

is this issue limited to one specific installation / computer, or does it occur with any installation of FreeBSD

Currently we have 2 FreeBSD servers.
On second tesseract runs without this issue. Versions there are:

tesseract 5.2.0
leptonica-1.82.0
libgif 5.2.1 : libjpeg 8d (libjpeg-turbo 2.1.3) : libpng 1.6.37+apng : libtiff 4.4.0 : zlib 1.2.11 : libwebp 1.2.4
Found OpenMP 201811
Found libarchive 3.6.1 zlib/1.2.11 liblzma/5.2.5 bz2lib/1.0.8 liblz4/1.9.3 libzstd/1.5.2
Found libcurl/7.85.0 OpenSSL/1.1.1o zlib/1.2.12 libpsl/0.21.1 (+libidn2/2.3.2) libssh2/1.10.0 nghttp2/1.48.0

Nothing about SSE here but than doesn's stop

Running inside vm used by https://www.hetzner.com/cloud
(dmesg indicates
CPU: AMD EPYC Processor (2445.48-MHz K8-class CPU
...
Hypervisor: Origin = "KVMKVMKVM"
...
ACPI APIC Table:
...
da0: <QEMU QEMU HARDDISK 2.5+> Fixed Direct Access SPC-3 SCSI device
)

[vmato]

Following trace may give additional info:
https://forums.freebsd.org/attachments/log_tail-txt.15069/

[amitdo]

The discussion in the FreeBSD forum indicates the issue is related to OpenMP.

OMP_THREAD_LIMIT=1
Didn't get any visible changes

It's not clear how you used this environment variable.

You need to do either of these:

OMP_THREAD_LIMIT=1 tesseract in.png output_name

Or

export OMP_THREAD_LIMIT=1
tesseract in.png output_name

[stweil]

According to your tesseract -v output, tesseract does not detect any SIMD instruction. not even SSE4.1.

That's a bug in Tesseract (fixed in commit adbefa8) which also caused bad performance for recognition and training on FreeBSD.

[vmato]

Looks very strange but now tesseract started working.

This happened before any software or OS updates.
So I can only suppose there was update of provider's virtualization software or something like that