Hi Justin, thank you for these bug reports.
> 1) getlo().unlink() doesn't raise an error if the LO doesn't exist,
> even though strace shows that libpq is returning an error.
This is caused by the problem you mention in your other mail - the
return value -1 is used for errors in lo_unlink(). A similar problem
existed in lo_export(). I have fixed this and added regression tests.
> 2) while debugging the first issue, I noticed that pgmodule.c:
> connGetLO() seems to stuff an Oid into an int. If I understand
> correctly that should be type:Oid, and use "OI" format ? I reported
> last year a similar issue involving pygres typecasting.
Made this a bit cleaner, but this shouldn't cause problem, since Oid is
defined as (unsigned) int in Postgres. The other issue had been fixed in
5.0.1 already.
> BTW, there's an unreleased change to pygres allowing
> query_formatted() without params. I wondered if that might be
> released soon.
+1 for cutting a 5.0.4 patch release. The changelog is already prepared,
only the release date needs to be inserted.
-- Christoph
_______________________________________________
PyGreSQL mailing list
[email protected]
https://mail.vex.net/mailman/listinfo.cgi/pygresql
