On Fri, 2008-04-11 at 00:17 +0200, Petter Reinholdtsen wrote: > > Could you test that for me and report the output on stderr (or part > > of it) and possibly valgrind output? > > It got rid of the uninitialised values issues reported by valgrind, > which is great. With this new version, I've been unable to reproduce > the crash, but as I said before, it is hard to trigger, so I am not > sure it is gone. > > I tried using this test script several times, without any luck getting > it to crash.
To give you an update I have replaced all of the group member list code to a more thoroughly tested set data structure. Also returned user and group names from the LDAP server are now checked to see if they are valid so no corruption with possibly binary data can take place any more (not 100% safe yet but should work). Running this code under valgrind does not show any memory problems in the nslcd code. > Note the 'nslcd: error writing to client' messages. Can these be a > problem? It seem to me that the server continue to work after the > client return. That can't be how it is supposed to work? If the request from the NSS client to nslcd times out from the client side the client closes the connection (see #474174). If the server has some data to flush but fails you will see this message. -- -- arthur - [EMAIL PROTECTED] - http://people.debian.org/~adejong --
signature.asc
Description: This is a digitally signed message part
