https://bugs.winehq.org/show_bug.cgi?id=45417
--- Comment #11 from Silvan s.jegen@gmail.com --- (In reply to Zebediah Figura from comment #10)
I suppose then windowscodecs isn't involved. I'm not sure what else would be used to read images (and I'm not sure what else could cause a discrepancy). I doubt I can get anything from a relay log, but I'll try taking a look at one anyway. If not perhaps someone else will know what to look for.
I uploaded the WINEDEBUG=+seh,+relay output here:
https://drive.google.com/file/d/1oKM97VFQ2zWSq1FY9IQcDSa2XNuCXC9X/view?usp=s...
Please let me know if you can think of anything else we could try.
I now tried another file containing not scans but jpgs resulting from converting an existing PDF. Those JPGs are very clean which means they shouldn't pose a big challenge for the OCR system. It turns out however that on Linux with Wine, while the text has been recognized correctly, some of the lines are missing the spaces between words for some reason. As far as I remember this is not an issue on Windows. I will attach the resulting .docx file and the original PDF file.