In the future, it might be good for ContentMine to start fuzzing its codebase. This would generate a more varied input than corpora do, and might help catch bugs that otherwise would not be caught, including potentially security issues.
LibreOffice - which, a bit like ContentMine, is designed to handle heterogeneous input files - has recently adopted fuzzing with some success: https://lwn.net/Articles/723566/
In the future, it might be good for ContentMine to start fuzzing its codebase. This would generate a more varied input than corpora do, and might help catch bugs that otherwise would not be caught, including potentially security issues.
LibreOffice - which, a bit like ContentMine, is designed to handle heterogeneous input files - has recently adopted fuzzing with some success: https://lwn.net/Articles/723566/