Since test008 worked nicely overnight I decided to get a bit more aggressive. I don't think it's related to connection code, but with HEAD (large batch of ADDs; local smoke tests, not from ldap-src tests):
t@16 (l@16) terminated by signal SEGV (no mapping at the fault address) Current function is __ham_func4 (optimized) 177 HASH4; current thread: t@16 =>[1] __ham_func4(dbp = ???, key = ???, len = ???) (optimized), at 0xffffffff7af16818 (line ~177) in "hash_func.c" [2] __db_chksum(hdr = ???, data = ???, data_len = ???, mac_key = ???, store = ???) (optimized), at 0xffffffff7af16c8c (line ~112) in "hmac.c" [3] __txn_force_abort(env = ???, buffer = ???) (optimized), at 0xffffffff7af43ad4 (line ~1653) in "txn.c" [4] __log_flush_commit(env = ???, lsnp = ???, flags = ???) (optimized), at 0xffffffff7af1f514 (line ~524) in "log_put.c" [5] __log_put(env = ???, lsnp = ???, udbt = ???, flags = ???) (optimized), at 0xffffffff7af1f0f8 (line ~297) in "log_put.c" [6] __txn_regop_log(env = ???, txnp = ???, ret_lsnp = ???, flags = ???, opcode = ???, timestamp = ???, envid = ???, locks = ???) (optimized), at 0xffffffff7af448a8 (line ~237) in "txn_auto.c" [7] __txn_commit(txn = ???, flags = ???) (optimized), at 0xffffffff7af41c2c (line ~687) in "txn.c" [8] __txn_commit_pp(txn = ???, flags = ???) (optimized), at 0xffffffff7af41790 (line ~539) in "txn.c" [9] hdb_add(op = 0x10adb77b0, rs = 0xffffffff42fff998), line 456 in "add.c" [10] overlay_op_walk(op = 0x10adb77b0, rs = 0xffffffff42fff998, which = op_add, oi = 0x1005dfd70, on = (nil)), line 670 in "backover.c" [11] over_op_func(op = 0x10adb77b0, rs = 0xffffffff42fff998, which = op_add), line 722 in "backover.c" [12] over_op_add(op = 0x10adb77b0, rs = 0xffffffff42fff998), line 768 in "backover.c" [13] fe_op_add(op = 0x10adb77b0, rs = 0xffffffff42fff998), line 334 in "add.c" [14] do_add(op = 0x10adb77b0, rs = 0xffffffff42fff998), line 194 in "add.c" [15] connection_operation(ctx = 0xffffffff42fffc20, arg_v = 0x10adb77b0), line 1100 in "connection.c" [16] connection_read_thread(ctx = 0xffffffff42fffc20, argv = 0x39), line 1226 in "connection.c" [17] ldap_int_thread_pool_wrapper(xpool = 0x1005352c0), line 663 in "tpool.c"
db4 4.7.25 patched for Oracle #16406, #16415, and #16541. All other slapd threads are idle. FWIW, I'm not fully patched for #5860 (will re-checkout now, and kill the optimization on libdb4-7).