Re: (ITS#5444) slapd seg. fault using mulitmirrored mode

[hyc@symas.com]

Pierangelo Masarati wrote:
> hyc@symas.com wrote:
>> ando@sys-net.it wrote:
>>> ando@sys-net.it wrote:
>>>> mohel@web.de wrote:
>>>>
>>>>> The above log was generated using the test050 script just adding a line which
>>>>> enters information also to a second server. Perhaps this test would also make
>>>>> sense for further releases.
>>>> Can you please post the modification to the script?
>>> Never mind, I could reproduce it.  It seems to be a duplicate of
>>> ITS#5437.  What actually happens is that if modifications are applied to
>>> all three servers, those applied to consumers do not persist, nor get
>>> replicated; on the contrary, the consumers crash later on as described
>>> in ITS#5437.
>> #5437? What you're talking about doesn't seem to match any text in ITS#5437.
>
> Well, sorry for not providing more details, I was going to investigate
> it a little bit more but I had to leave.  The crash to me occurs in
> syncprov_done_ctrl(), as indicated in ITS#5437.  What happens is that
> syncprov_done_ctrl() with an invalid cookie (unset; clearing it in
> syncprov_search_response(), as suggested in ITS#5437 leads to the
> behavior I described (writes to consumers are simply ignored, don't
> persist in the consumers and are not replicated).  To reproduce, add
> something to a consumer close to the end of test050, before killing the
> servers, and wait a little bit to let things proceed to the crash.

OK, I've reproduced the crash and patched it in HEAD, and added a consumer 
write to the test050 script. Please test.

-- 
   -- Howard Chu
   CTO, Symas Corp.           http://www.symas.com
   Director, Highland Sun     http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP  http://www.openldap.org/project/