[PATCH] replmd changes to handle single-valued link conflicts

[Garming Sam]

Hi Tim,

- For the function replmd_check_singleval_la_conflict, can you break 
down the comment a bit? It's a large block of text that I'd probably 
just skip over.

- In the commit message for replmd: Fix RMD_VERSION initial value to 
match Windows, can you cite the sub-heading and section numbers for the 
MS-DRSR spec? It's just one more clue that would be useful.

- In replmd: Make replmd_set_la_val() closer to replmd_build_la_val(), 
the assignment of old_dsdb_dn to NULL should have been with the original 
code. Perhaps this patch should occur at the beginning.

- Bug numbers need to be on the later patches


So if an incoming link is newer, it will always clobber any existing 
links regardless of either being active or inactive?

The overall patches seem pretty reasonable, I think some changes could 
have been squashed in a bit more, but I don't think it matters.

I noticed you mentioned in 
https://bugzilla.samba.org/show_bug.cgi?id=13039 that it looks like the 
object GUID is being used rather than the invocation ID, did you check 
if that was the case here? Did we just never bother to implement the 
actual object conflict algorithm at all?


Cheers,

Garming


On 28/09/17 17:01, Tim Beale via samba-technical wrote:
> Hi,
>
> The attached patch-set fixes the following bugs:
> - 13055: Samba replication cannot resolve single-valued link conflicts
> - 13059: Link attributes on Samba have different initial RMD_VERSION
> compared to Windows
>
> The bulk of the changes are to handle conflicts correctly for
> single-valued link attributes. It required a bit of refactoring in the
> replmd code to support this.
>
> Patches can also be viewed here:
> http://git.catalyst.net.nz/gw?p=samba.git;a=shortlog;h=refs/heads/tim-replmd-latest
>
> Please let me know if you have any review comments.
>
> Thanks,
> Tim
>