SIGSEGV from START_REPLICATION 0/XXXXXXX in XLogSendPhysical () at walsender.c:2762

Pgjdbc test suite identified a SIGSEGV in the recent HEAD builds of PostgreSQL, Ubuntu 14.04.5 LTS

Here's a call stack: https://travis-ci.org/github/pgjdbc/pgjdbc/jobs/691794110#L7484

The crash is consistent, and it reproduces 100% of the cases so far.

The CI history shows that HEAD was good at 11 May 13:27 UTC, and it became bad by 19 May 14:00 UTC,

so the regression was introduced somewhere in-between.

Does that ring any bells?

In case you wonder:

Program terminated with signal SIGSEGV, Segmentation fault.
#0  XLogSendPhysical () at walsender.c:2762
2762 if (!WALRead(xlogreader,
(gdb) #0  XLogSendPhysical () at walsender.c:2762
        SendRqstPtr = 133473640
        startptr = 133473240
        endptr = 133473640
        nbytes = 400
        segno = 1
        errinfo = {wre_errno = 988942240, wre_off = 2, wre_req = -1,
          wre_read = -1, wre_seg = {ws_file = 4714224,
            ws_segno = 140729887364688, ws_tli = 0}}
        __func__ = "XLogSendPhysical"
#1  0x000000000087fa8a in WalSndLoop (send_data=0x88009d <XLogSendPhysical>)
    at walsender.c:2300
        __func__ = "WalSndLoop"
#2  0x000000000087d65a in StartReplication (cmd=0x299bda8) at walsender.c:750
        buf = {data = 0x0, len = 3, maxlen = 1024, cursor = 87}
        FlushPtr = 133473640
        __func__ = "StartReplication"
#3  0x000000000087eddc in exec_replication_command (
    cmd_string=0x2916e68 "START_REPLICATION 0/7F4A3D8") at walsender.c:1643
        cmd = 0x299bda8
        parse_rc = 0
        cmd_node = 0x299bda8
        cmd_context = 0x299bc60
        old_context = 0x2916d50
        qc = {commandTag = 988942640, nprocessed = 140729887363520}
        __func__ = "exec_replication_command"