Port winbindd child dispatch table split up to v3-2-test

[Stefan (metze) Metzmacher]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Gerald (Jerry) Carter schrieb:
> Metze,
> 
>> - Convert all internal communication to a NDR-based protocol.
>>   And the different dispatch tables would allow be to convert
>>   each on its own to the new protocol. While this also trims down
>>   the test matrix, so I can be sure I don't need to test e.g.
>>   the WINBINDD_DSGETDCNAME operation in a domain or idmap child.
>>   For me it makes it much easier to understand the scope of
>>   the operations.
> 
> So if I understand correctly, you will have 3 IDLs (one for each
> table).  What happens if I need a new function in the idmap child
> and in a domain child.  I now have to add it to two IDL files?
> Not a deal killer I guess.  Just trying to make sure I understand
> the ramifications of splitting the function table up.

Hi Jerry,

No, we only have 1 IDL file, there's no relation between numbers of
child dispatch tables and idl interfaces.

But not each child will response to each function, so the e.g. domain
child will directly return with an error or panic if it gets a
WINBINDD_DSGETDCNAME request, as that means the programmer gets
something wrong in the parent code, as this request should never reach a
domain child. And if a domain child would go into the
WINBINDD_DSGETDCNAME codepath, it's likely that it will crash or
something strange anyway, but this would be happen randomly and it's
hard to debug, so I think it's better to just give an "not implemented"
error or something like this.

metze
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iD8DBQFHFxSPm70gjA5TCD8RAgAHAKDKL/QLVyWb2i2T1PDt1KNV61RFQQCdGNBN
9qZ7bm2IIoJXfSxQw+bgMk8=
=e0k0
-----END PGP SIGNATURE-----