Currently, darshan logutils API expects an input buffer to be passed in to store module records it reads from the log file. The problem is that the caller might not always know how big the record it is about to extract is, particularly in cases where a module's records are variable-length (in which case one would need to sift through the data to determine the record size).

It would probably be better to just have the logutils API malloc memory for the current buffer it is extracting rather than relying on the caller to ensure there is enough space. This buffer can then be returned to the user where it will eventually need to be freed up.