On 9/27/07, Jan Wieck <JanWieck at yahoo.com> wrote:
> On 9/26/2007 6:29 AM, Marko Kreen wrote:
> > What now?  We should run this thru -hackers too.  I could
> > even prepare patch for /contrib, to show the code too and maybe
> > the hell will froze for a moment and this will get into 8.3 ;)
>
> I've talked to Tom briefly on -core. What we need to do is to agree on a
> module name, the naming conventions (I think we got those already) and
> get the final code done within the next 2 weeks. I'll be in Hong Kong
> October 10-11, so that week is almost gone with meetings, travel and
> fighting jet lag. I will review the code when I'm back, test it using
> Slony-I HEAD by actually porting Slony to it, and add it to contrib.

Great news!

Here is first draft of the contrib/txid then.

Final API:

int8          txid_current()
txid_snapshot txid_current_snapshot()
int8          txid_snapshot_xmin(snap)
int8          txid_snapshot_xmax(snap)
setof int8    txid_snapshot_xip(snap)
bool          txid_visible_in_snapshot(int8, snap)
bool          txid_not_visible_in_snapshot(int8, snap)


Changes to the code in Skytools:

- renamed the functions
- dropped the epoch-in-table code
- tagged the txid_current/txid_current_snapshot as STABLE
  (they were SECURITY DEFINERs because of the epoch-in-table)
- simplified the txid_snapshot_xip by use of fctx->call_cnt
- removed use of "unsigned long long" in txid_snapshot_out,
  instead I try to guess the correct fromat string for uint64
  from HAVE_LONG_INT_64

TODO:
- flesh out the README
- review/testing by somebody who knows arches with strict
  alignment requirements.
- some more testing, to see if i did not broke anything...

-- =

marko
-------------- next part --------------
A non-text attachment was scrubbed...
Name: txid-v1.diff.gz
Type: application/x-gzip
Size: 6957 bytes
Desc: not available
Url : http://lists.slony.info/pipermail/slony1-hackers/attachments/20070928=
/1770a8b3/txid-v1.diff.bin