CAUSING MYSQL TO STALL

Problem:

When GTID_MODE=OFF, calling "PURGE BINARY LOGS TO '<FILE>'" causes the
system to read the whole '<FILE>' file. When the referenced file is
big, this may stall the server for a few seconds.

Analysis:

The server needs to read the Previous_gtids of the first binary log
file kept after the purge in order to compute the GTID_PURGED.

There is an optimization that stops the reading in the binary log
file once finding the first GTID event. That's why the issue is
only happening when GTID_MODE=OFF.

Fix:

Added an optimization to stop reading the binary log file when
it is supposed to read only the "lost" GTIDs and a PREVIOUS_GTIDS
event is found. This will cover the default 5.7.7+ behavior because
of binlog_gtid_simple_recovery option.