On Fri, Dec 15, 2017 at 10:58 AM, Andres Freund <andres@anarazel.de> wrote:
>> I have one minor piece of feedback on the upgrading of assertions to
>> ereport()s with ERRCODE_DATA_CORRUPTION: It would be nice if you could
>> upgrade the raw elog() "can't happen" error within
>> IndexBuildHeapRangeScan() to be an ereport() with
>> ERRCODE_DATA_CORRUPTION. I'm referring to the "failed to find parent
>> tuple for heap-only tuple" error, which I think merits being a real
>> user-visible error, just like the relfrozenxid/relminmxid tests are
>> now. As you know, the enhanced amcheck will sometimes detect
>> corruption due to this bug by hitting that error.
>
> I'm not opposed to that, it just seems independent from this thread. Not
> sure I really want to go around and backpatch such a change, that code
> has changed a bit between branches. Happy to do so on master.
The elog(), which was itself upgraded from a simple Assert by commit
d70cf811, appears in exactly the same form in 9.3+. Things did change
there, but they were kept in sync.
--
Peter Geoghegan
