There is a bug in FuzzyOCR where it keeps reusing the some tmpdir for images, so processes step on each other's stuff, resulting in errors like this in the log: Error running preprocessor(pamthreshold): /usr/bin/pamthreshold -simple -threshold 0.5 pamthreshold: bad magic number - not a PAM, PPM, PGM, or PBM file See http://fuzzyocr.own-hero.net/ticket/421 and http://fuzzyocr.own-hero.net/ticket/110 Also, there is another bug where .'s are used in the image file name when it is generated from the content-id, causing an error when the part after the . is misinterpreted as a file extension. See http://fuzzyocr.own-hero.net/ticket/60 Please consider including the attached patches to fix these bugs.
Created attachment 175916 [details, diff] Patch for tmpdir bug
Created attachment 175917 [details, diff] Patch for . in file name bug
Created attachment 195701 [details] Overlay with all patches this overlay is build with all patches I can found for FuzzyOcr
Thanks for the report! And sorry for the delay I've just added 3.6.0 in tree, with both patches included (they were reported separately in other bugreports)