On Fri, Jan 25, 2013 at 05:30:58PM -0500, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > OK, updated patch attached that throws an error with a more specific
> > message.
>
>
> > *
> > * As noted above rd_newRelfilenodeSubid is not set in all cases
> > * where we can apply the optimization, so in those rare cases
> > ! * where we cannot honor the request.
> > */
>
> This sentence not complete. I kind of think the entire para visible
> above could be removed, anyway.
>
> > ! ereport(ERROR, (errmsg("cannot perform FREEZE operation due to invalid table or transaction
state")));
>
> I don't find this terribly specific. It would at least be useful to
> have two messages distinguishing whether the cause was invalid table
> state (rd_createSubid and rd_newRelfilenodeSubid not set) or invalid
> transaction state (the snapshot and portal tests). The former might
> usefully be phrased as "because the table was not created or truncated
> in the current transaction" and the latter as "because other actions are
> in progress within the current transaction".
>
> I'd also suggest "cannot perform COPY FREEZE because <whatever>" rather
> than using the unnecessarily vague "operation".
>
> Also, this is missing an errcode, which means it will report itself as
> an internal error, which it ain't. It's also randomly unlike the
> standard layout for ereport calls.
> ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE would do for the table case,
> not sure about the other.
OK, that was tricky, but completed with the attached patch.
Surprisingly, truncation wasn't mention in our docs, though it was used
in the regression tests. I have fixed that.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +