On Sat, Aug 31, 2013 at 11:23 AM, Devaraj Das <[email protected]> wrote:
> Thanks for making the RC, Stack. I bumped into one thing which I thought I
> should bring up in the context of singularity - the ServerName message in
> hbase.proto should have the start_code as a signed int. We allow for -1 (
> ServerName.NON_STARTCODE) as a start_code. Hence.. Yes it can be worked
> around, and, maybe there won't any wire-compat issues if we make the change
> later (not sure about it) but I think we should fix it now. What do you
> think?
>
> For illustration, here is the proposed fix.
>
> diff --git a/hbase-protocol/src/main/protobuf/hbase.proto
> b/hbase-protocol/src/main/protobuf/hbase.proto
> index 08061e5..2f9a8d1 100644
> --- a/hbase-protocol/src/main/protobuf/hbase.proto
> +++ b/hbase-protocol/src/main/protobuf/hbase.proto
> @@ -120,7 +120,7 @@ enum CompareType {
> message ServerName {
> required string host_name = 1;
> optional uint32 port = 2;
> - optional uint64 start_code = 3;
> + optional int64 start_code = 3;
> }
>
>
I'd say it not a deal breaker; more a good to fix in the next RC if there
is one (my guess is that there will likely be one -- smile). I
made HBASE-9408 as a placeholder for now.
Thanks for taking a look DD,
St.Ack
