RE: [nssldap] More patches to nss_ldap 265
[
Date Prev][
Date Next]
[
Thread Prev][
Thread Next]
RE: [nssldap] More patches to nss_ldap 265
- From: "Howard Wilkinson" <howard [at] cohtech.com>
- To: "Douglas E. Engert" <deengert [at] anl.gov>
- Cc: <nssldap [at] padl.com>
- Subject: RE: [nssldap] More patches to nss_ldap 265
- Date: Wed, 9 Dec 2009 09:24:25 -0000
Doug,
Howard Wilkinson wrote:
> Sorted this one out I think. Added a mapping routine going
back from ns_ldap_return_code to NSS_STATUS and called that for all returns
from parseEntry, which turns out to be an internal function - could be declared
static probably....
>
> Now run into another problem. The Solaris documentation
defines the spwd (shadow password) structure as being populated with long's,
which corresponds with that used by Linux. However, the header declares these
to be ints.
>
> I am going to fix this with conditional compilation keyed
from __solaris__ unless someone can suggest a better solution.
>
Is this a problem? Why do you need to change this. It could
have implications for 64 bit.
Yes it is as the port I am building for is 64-bit and the definitions are 'int'
not long so although this would work by mistake on a 64-bit machine it would
fail on a 32-bit machine with interesting behaviour that is architecture
dependent i.e. x86 would give a different result from the sparc platform. Have
now coded this to use __sun as the conditional and got it compiling - need to
test my solution but it looks reasonable (and not too inelegant)
> Coherent Technology Limited, 23 Northampton Square, Finsbury,
London EC1V 0HL, United Kingdom
> Telephone: +44 20 7690 7075 Mobile: +44 7980 639379
> Company Email: coherent@cohtech.com Website:
http://www.cohtech.com <http://www.cohtech.com/> <http://www.cohtech.com/>
>
> ________________________________
>
> From: owner-nssldap@padl.com on behalf of Howard Wilkinson
> Sent: Tue 2009-12-08 15:28
> To: nssldap@padl.com
> Subject: RE: [nssldap] More patches to nss_ldap 265
>
>
>
> I am in the process of porting the code over to Solaris and
have come across a section of code I have never looked at within ldap-sldap.c.
The code is the externally available function __ns_ldap_parseEntry which is
declared to return an NSS_STATUS but is returning either an NSS_STATUS or a
ns_ldap_return_code.
>
> This function is not used anywhere else in the library. From
static analysis of the code it looks like it can never have worked properly as
it is returning error results that would look pseudo-random. Does anybody know
anything about this and if it is a really essential part of the library. If it
is then I need to fix and could do with some advice on the correct
interface/return codes. If it is not I will fix as best I can see fit and mark
this for removal in a later patch set.
>
> Coherent Technology Limited, 23 Northampton Square, Finsbury,
London EC1V 0HL, United Kingdom
> Telephone: +44 20 7690 7075 Mobile: +44 7980 639379
> Company Email: coherent@cohtech.com Website:
http://www.cohtech.com <http://www.cohtech.com/> <http://www.cohtech.com/>
<http://www.cohtech.com/>
>
> ________________________________
>
> From: owner-nssldap@padl.com on behalf of Howard Wilkinson
> Sent: Mon 2009-12-07 07:52
> To: nssldap@padl.com
> Subject: [nssldap] More patches to nss_ldap 265
>
>
>
> I have just uploaded 3 more patches to the nss_ldap 265 code.
These have all been attached to the bug #412.
>
> The patches deal with reconnection timeouts and SRV record
sorting.
>
> 1.
> A simple change to make the logic handling
maxsleeptime work as originally intended. The change in behaviour is to enforce
the maxsleeptime option as a barrier to the increase of the backoff time rather
than as at present where it the backoff could grow to almost twice the
maxsleeptime.
> 2.
> A change to increase the accuracy of the sleep
periods - the system can now be configured down to nanosecond level - although
the actual accuracy will depend on the systems ability to handle such durations
accurately. A further modification in this patch adds alternative backoff
growth algorithms so that the exponential backoff growth is now a default
rather than the only option.
> 3.
> Code in the dnsconfig area that allow the sorting of
the SRV records by priority and weight. This allows data centre configuration
of complex load balancing strategies.
>
> If anybody has the time to try these out (along with my other
patches) and report back I would be grateful for the feedback. Also, if anybody
has a working Solaris 10 build for this code (with or without my patches) could
they send me the details as I need to make sure my changes will build and
operate on the Solaris platform.
>
> Regards, Howard
>
> Coherent Technology Limited, 23 Northampton Square, Finsbury,
London EC1V 0HL, United Kingdom
> Telephone: +44 20 7690 7075 Mobile: +44 7980 639379
> Company Email: coherent@cohtech.com
<coherent [at] cohtech.com> Website: http://www.cohtech.com
<http://www.cohtech.com/> <http://www.cohtech.com/> <http://www.cohtech.com/>
<http://www.cohtech.com/>
>
>
>
>
>
--
Douglas E. Engert <DEEngert@anl.gov>
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
(630) 252-5444