From: Tom Lane <tgl@sss.pgh.pa.us>
To: Nathan Bossart <nathandbossart@gmail.com>
cc: Martin Kalcher <martin.kalcher@aboutsource.net>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: [Patch] Fix bounds check in trim_array()
In-reply-to: <20220728224618.GA412680@nathanxps13>
References: <baaeb413-b8a8-4656-5757-ef347e5ec11f@aboutsource.net>
 <20220728224618.GA412680@nathanxps13>
Comments: In-reply-to Nathan Bossart <nathandbossart@gmail.com>
	message dated "Thu, 28 Jul 2022 15:46:18 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1233970.1659288333.1@sss.pgh.pa.us>
Date: Sun, 31 Jul 2022 13:25:33 -0400
Message-ID: <1233971.1659288333@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1233971.1659288333%40sss.pgh.pa.us>
Precedence: bulk

Nathan Bossart <nathandbossart@gmail.com> writes:
> On Mon, Jul 25, 2022 at 04:40:51PM +0200, Martin Kalcher wrote:
>> +SELECT trim_array(ARRAY[]::int[], 1); -- fail
>> +ERROR:  number of elements to trim must be between 0 and 0

> Can we improve the error message?  Maybe it should look something like
> 	ERROR:  number of elements to trim must be 0
> for this case.

Hmm, I'm unexcited about making our long-suffering translators
deal with another translatable string for such a corner case.
I think it's fine as-is.

A bigger problem is that a little further down, there's an equally
unprotected reference to ARR_LBOUND(v)[0].  Now, the fact that that
expression computes garbage doesn't matter too much, because AFAICS
if the array is zero-D then array_get_slice is going to exit at

	if (ndim < nSubscripts || ndim <= 0 || ndim > MAXDIM)
		return PointerGetDatum(construct_empty_array(elemtype));

without ever examining its upperIndx[] argument.  However,
once we put in a test case covering this behavior, I bet that
valgrind-using buildfarm animals will start to bleat about the
invalid memory access.  I think the easiest fix is like

	if (ARR_NDIM(v) > 0)
	{
		upper[0] = ARR_LBOUND(v)[0] + array_length - n - 1;
		upperProvided[0] = true;
	}

It'd be good to get this fix into next week's minor releases,
so I'll go push it.

			regards, tom lane