While testing out the final feature yet to be released, I noticed that archives were being created that were slightly larger than the max size allowed. i.e. This bug entirely defeats the purpose of the program.
I need feedback from people using tarlimit. My tests run on a small number of files with size limits less than 10M. Is this bug occurring when the size limit is in the range of Gigabytes? Is this bug occurring only when files are small, or when they are large as well?
In the meantime, I'm going to be running some tests on two theories. The first is that size() isn't reporting correct numbers. The second is that tar.add() does not block, so the program goes on and reads the size before the previous item is finished being written.
Logged In: YES
user_id=930755
Originator: YES
I've narrowed the problem down. Using os.path.getsize() on an open file apparently isn't reliable. To make matters more fun, a TarFile object can't be flushed, nor can it be closed and reopened for appending if it is a compressed tar.