More cleanup since yesterday. I found that some SVGs have bogus linked images, which makes inkscape show a big red cross saying “linked image not found”. I added a pre-filtering stage before the thumbnails are created in which the SVG is cleaned up. In that process, bogus image insertions are deleted, and non-ASCII characters are replaced by spaces. This is another issue that appeared with non us-english files. I found that there are characters over 127, but the XML file is assumed to be ascii, which again makes inkscape and other programs I’m using throw errors.
I ran the thumbnail regeneration script and lots of images are getting cleaned up. There are still some problems with ODG export, and little by little I’ll get all of the problems taken care of.