- # we don't know where we're going to save this stuff yet,
- # so I guess we'll write it to another tempfile. One we know the path of.
- # I'm assuming the tempfile we get from pylons is set to delete itself
- # when it closes, and has no visible path. Maybe I'm wrong?
- intermediate_file_descriptor, intermediate_path = tempfile.mkstemp()
+ # The incoming fileobj could be a tempfile that's already been unlinked --
+ # and probably is, as it's coming from a Pylons upload object. Thus we
+ # have to copy the data rather than the file, and we may have to read the
+ # file anyway to hash it, so do both at the same time.