Alexander Kobel

Sounds good. In this case, I'll remove the octave bindings from the Arch Linux build, too. Thanks for the investigation!

Thanks @juandelperal for your quick response. Maybe you can clarify one more thing: As far as I can see (no font expert here, just an avid user), - both TTF...

For the sake of easier access: ``` #!/bin/sh tesseract --version echo wget -qN https://www.adobe.io/content/dam/udp/en/open/standards/tiff/TIFF6.pdf mutool draw -q -c mono -o %d.pbm -r300 TIFF6.pdf 1,2 for page in 1 2; do...

Reverting 2881dfb049aea0821b506e5a5ed0048eef749c04 resolves the issue.

Disclaimer: not much of PDF expertise here. I'm much more surprised about the difference on the second page for the multi-page msb-to-lsb min-is-white. [edit: not anymore, this is simply twice...

Also, the PDFs created by tesseract from the multi-page TIFFs have twice the first page. The image itself seems intact. Not digging deep in the code, but (how) does `extractG4DataFromFile`...

There is one pretty obvious and relevant source for images where this should work without much hassle: ones that are generated by extracting them from valid PDFs. Not sure of...

In contrast, by the way, to lossy compression - where re-encoding can be actually harmful not just from an efficiency point of view...

@DanBloomberg Works fine. [edit: Tested on current master HEAD 2f2e488, which includes both 0487bc5 and b68c656.] Note that I only tested the script from the initial post and a couple...

Thanks @DanBloomberg for considering this issue and producing a fix so swiftly! I reported a bug for the Arch Linux leptonica package at https://bugs.archlinux.org/task/71856 when I realized the problem, referencing...