On 10/1/07, Tom Lane <tgl at sss.pgh.pa.us> wrote:
> "Marko Kreen" <markokr at gmail.com> writes:
> > I don't understand you.  Current code _is_ making them
> > epoch-independent?  What aspect of it do you dislike?
>
> InvalidTransactionId should compare as less than any other XID;
> unless you are doing something very strange, that won't happen
> with this coding.

Ah, then it's indeed my mistake.  I think I saw some code doing
if (!TransactionIdIsValid()) and assuming invisiblility, but
now I look the comparision functions (TransactionIdPrecedes, etc)
they indeed operate as you say.

Attaches is v5, with following patch compared to v4:

@@ -96,10 +96,6 @@ convert_xid(TransactionId xid, const TxidEpoch *state)
 {
        uint64 epoch;

-       /* avoid issues with the the special meaning of 0 */
-       if (xid =3D=3D InvalidTransactionId)
-               return MAX_TXID;
-
        /* return special xid's as-is */
        if (xid < FirstNormalTransactionId)
                return xid;

-- =

marko
-------------- next part --------------
A non-text attachment was scrubbed...
Name: txid-v5.diff.gz
Type: application/x-gzip
Size: 6617 bytes
Desc: not available
Url : http://lists.slony.info/pipermail/slony1-hackers/attachments/20071001=
/9154ca52/txid-v5.diff.bin