Incomplete Binlog writes on disk-full error, also breaking slave replication

It appears the binlog and replication code still has write calls that aren't checked for successful completion. On disk full, the binlog gets truncated.

As a direct consequence and example, a slave will report this error:
Got fatal error 1236 from master when reading data from binary log: 'binlog truncated in the middle of event'

Obviously it's not just a problem with replication, but also for having correct binlogs for point-in-time recovery.

Suggested solution: writes should check for success, and in case of a disk-full error, report and retry correctly. The "wait until space has been freed up" approach that's used on other disk-full conditions, would be fine here also.