Bug#816294: slapd segfault on update dbconfig

[Ryan Tandy]

Control: tag -1 upstream

On Thu, Mar 10, 2016 at 09:22:43PM -0800, Ryan Tandy wrote:

It looks non-trivial...

That is, it's not just a matter of a missing NULL check - I'm not sure what
syncprov_checkpoint is supposed to do in this situation.

0x00007fffefac8929 in hdb_modify (op=0x7fffffffde00, rs=0x7fffffffdd90) at 
modify.c:555

Same thing with upstream git sources, RE24 and master, on unstable.

0x00000000004e2408 in hdb_modify (op=0x7fffffffd9c0, rs=0x7fffffffd950) at 
modify.c:522
522                     rs->sr_err = TXN_BEGIN( bdb->bi_dbenv, NULL, &ltid, 
tflags );
(gdb) bt
#0  0x00000000004e2408 in hdb_modify (op=0x7fffffffd9c0, rs=0x7fffffffd950) at 
modify.c:522
#1  0x000000000053bddd in syncprov_checkpoint (op=0x7fffffffdeb0, on=0x90e390) 
at syncprov.c:1495
#2  0x0000000000541910 in syncprov_db_close (be=0x9088d0, cr=0x0) at 
syncprov.c:3237
#3  0x00000000004ba0e8 in over_db_close (be=0x9088d0, cr=0x0) at backover.c:180
#4  0x000000000043ca4d in backend_shutdown (be=0x9088d0) at backend.c:383
#5  0x0000000000468d06 in slap_shutdown (be=0x0) at init.c:233
#6  0x0000000000405bba in main (argc=9, argv=0x7fffffffe548) at main.c:1028

I do not see it with 2.4.31, though, so that gives me a basis to try 
bisecting...