From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Joe Conway <mail(at)joeconway(dot)com> |
Cc: | MauMau <maumau307(at)gmail(dot)com>, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: [bug fix] Memory leak in dblink |
Date: | 2014-06-19 03:19:57 |
Message-ID: | 11929.1403147997@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Joe Conway <mail(at)joeconway(dot)com> writes:
> On a side note, while perusing this section of code:
> 8<-------------------------- at dblink.c:1176 --------------------------
> /* make sure we have a persistent copy of the tupdesc */
> tupdesc = CreateTupleDescCopy(tupdesc);
> Shouldn't that CreateTupleDescCopy() happen in ecxt_per_query_memory?
Not necessary (we'd have seen crashes long since if it was).
ExecMakeTableFunctionResult doesn't need the tupdesc to persist past
return.
Actually, I was wondering whether we couldn't remove that
CreateTupleDescCopy call entirely. The risk would be if
get_call_result_type returned a pointer into relcache or some other cached
tuple descriptor, which might be subject to a cache flush --- but AFAICS
it always returns a freshly created or copied tupdesc. (This might not
have been true originally, which could explain why dblink thinks it needs
to copy.)
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Peter Geoghegan | 2014-06-19 03:30:07 | Re: Possible index issue on 9.5 slave |
Previous Message | Ian Barwick | 2014-06-19 03:09:22 | Re: Possible index issue on 9.5 slave |