[PATCH] RFC debug detect winbindd version mismatches

[Noel Power]

On 27/01/14 13:18, Volker Lendecke wrote:
> On Mon, Jan 27, 2014 at 07:27:34AM -0500, Jim McDonough wrote:
>> On Mon, Jan 27, 2014 at 6:20 AM, Volker Lendecke
>> <Volker.Lendecke at sernet.de> wrote:
>>> On Mon, Jan 27, 2014 at 10:35:47AM +0000, Noel Power wrote:
>>>> A simple debug related tweak that adds a new wbcError category which
>>>> allows winbindd version mismatch errors to be detected/reported. Its
>>>> likely one of the many wbcErrorString around the code will now print out
>>>> some more useful info. Please review/push if you think it's useful
>>> libwbclient is supposed to be shipped with winbind, it
>>> should always come in the same RPM or have a strict version
>>> dependency. In what circumstances do you need better
>>> diagnostics for this?
>> It's not so much _needed_ as helpful.  If a loaded binary, such as
>> sshd, isn't restarted, it's could be a quicker or more obvious
>> indicator.  Or in the case where an administrator has done something
>> unwise like manually installing binaries/moving things around.  Not
>> strictly needed, but could be a time saver.
>>
>> Not worth a lot of effort/comment, but if someone else found it useful
>> for their support folks (ours cover a wide range of topics) to guide
>> their thinking...
> I'm not opposed to this feature. Question: Is there a way to
> do this without a global variable?
everything is possible :-)
>  Maybe with a special call
> that queries the currently available interface version
unfortunately 'winbind_open_pipe_sock' has a version request as part of
the connection setup, if the version check fails then then you have no
chance to get the result of a subsequent version request (the way the
code currently is written)
> , or
> some other error code? I know that NSS_STATUS is a given,
> maybe we should pass something else here?
the NSS_STATUS errors are returned further up the stack ( e.g. after a
successful winbind_open_pipe_sock has returned ) as a result of
subsequent send/receive request/responses functions being called
>  Just
> brainstorming...
yup, understood, wasn't completely in love with the solution myself but
seemed the easiest less impactful way to achieve this (mostly debug
relevant) change, I can rewrite this by changing
winbind_open_pipe_sock/winbind_read_sock/winbind_write_sock (and
associated call chain) if the consensus is that it is necessary


Noel