Just after cutting 1.6.5, the unit test suite failed in test_backup03.
The failure was here (<http://mjc.homeunix.org:8180/job/wiredtiger-test-unit/2012/console>):
(gdb) up 4
WT-4 0x00007f08cd5f0f2e in __clsm_put (session=0x3c6ced0, clsm=0x5eb2b00, key=
0x5eb2bb8, value=0x5eb2bd8, position=0) at ../src/lsm/lsm_cursor.c:1006
1006 WT_ASSERT(session,
(gdb) where
#0 0x0000003fe9a35c55 in __GI_raise (sig=sig@entry=6)
at ../nptl/sysdeps/unix/sysv/linux/raise.c:63
WT-1 0x0000003fe9a37408 in __GI_abort () at abort.c:90
WT-2 0x00007f08cd5fd0a1 in __wt_abort (session=0x3c6ced0)
at ../src/os_posix/os_abort.c:24
WT-3 0x00007f08cd619dd0 in __wt_assert (session=0x3c6ced0, error=0, file_name=
0x7f08cd62ae8e "../src/lsm/lsm_cursor.c", line_number=1007, fmt=
0x7f08cd62ae8b "%s") at ../src/support/err.c:408
WT-4 0x00007f08cd5f0f2e in __clsm_put (session=0x3c6ced0, clsm=0x5eb2b00, key=
0x5eb2bb8, value=0x5eb2bd8, position=0) at ../src/lsm/lsm_cursor.c:1006
WT-5 0x00007f08cd5f16aa in __clsm_insert (cursor=0x5eb2b00)
at ../src/lsm/lsm_cursor.c:1119
WT-6 0x00007f08cd5e8d33 in __curtable_insert (cursor=0x5eb2f60)
at ../src/cursor/cur_table.c:380
(gdb) print session->txn.id
$1 = 154453
(gdb) print clsm->primary_chunk->txnid_max
$2 = 0
(gdb) print *clsm->primary_chunk
$3 = {id = 10, generation = 0, uri =
0x7f085c008f30 "file:test_backup.4_cgroup1-000010.lsm", bloom_uri = 0x0,
count = 0, create_ts = {tv_sec = 1381291510, tv_nsec = 674578609},
refcnt = 0, txnid_max = 0, flags = 0}
This shouldn't happen. It looks like either:
1. the cursor has switched to a new chunk, but the code in __clsm_enter to bump txnid_max hasn't worked; or
2. some thread has switched txnid_max to zero.
So far, I haven't been able to reproduce this.