Re: SerializeParamList vs machines with strict alignment - Mailing list pgsql-hackers

On Tue, Oct 2, 2018 at 11:21 AM Amit Kapila <amit.kapila16@gmail.com> wrote:
>
> On Tue, Oct 2, 2018 at 10:07 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> >
> > Amit Kapila <amit.kapila16@gmail.com> writes:
> > > On Tue, Oct 2, 2018 at 9:38 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > >> (But it might be worth choosing slightly less
> > >> generic object names, to avoid a conflict against other sub-tests
> > >> later in that script.)
> >
> > > The function name and statement name seems okay to me.  How about
> > > changing the table name to fooarr or arrtest?
> >
> > Up to you.
> >
>
> Okay, I have pushed the test case patch on HEAD.  Attached is the
> code-fix patch, let's wait for a day so that we have all the results
> which can help us to discuss the merits of this patch.
>

By now, the added test has failed on gharial [1] with below log on the server:
2018-10-02 00:46:57.019 MDT [5bb31455.2066:193] LOG:  statement:
CREATE TABLE fooarr(f1 text, f2 int[], f3 text);
2018-10-02 00:46:57.147 MDT [5bb31455.2066:194] LOG:  statement:
INSERT INTO fooarr VALUES('1', ARRAY[1,2], 'one');
2018-10-02 00:46:57.149 MDT [5bb31455.2066:195] LOG:  statement:
PREPARE pstmt(text, int[]) AS SELECT * FROM fooarr WHERE f1 = $1 AND
f2 = $2;
2018-10-02 00:46:57.153 MDT [5bb31455.2066:196] LOG:  statement:
EXPLAIN (COSTS OFF) EXECUTE pstmt('1', make_some_array(1,2));
2018-10-02 00:46:57.157 MDT [5bb31455.2066:197] LOG:  statement:
EXECUTE pstmt('1', make_some_array(1,2));
2018-10-02 00:46:57.157 MDT [5bb31455.2066:198] DETAIL:  prepare:
PREPARE pstmt(text, int[]) AS SELECT * FROM fooarr WHERE f1 = $1 AND
f2 = $2;
2018-10-02 00:46:57.784 MDT [5bb313c8.1da5:4] LOG:  server process
(PID 8294) was terminated by signal 10
2018-10-02 00:46:57.784 MDT [5bb313c8.1da5:5] DETAIL:  Failed process
was running: EXECUTE pstmt('1', make_some_array(1,2));
2018-10-02 00:46:57.784 MDT [5bb313c8.1da5:6] LOG:  terminating any
other active server processes

AFAICS, the machine details are as follows:
uname -m = ia64
uname -r = B.11.31
uname -s = HP-UX
uname -v = U

This seems to indicate that this hardware is alignment-sensitive.  I
haven't done a detailed study of this architecture, but if anybody is
familiar with this architecture, then feel free to share your
thoughts.  On a quick look at one of on this architecture, there
doesn't seem to be any recommendation on aligning the memory access to
8-byte boundary.  See 3.2.2 section (Addressable Units and Alignment)
in doc [2].

This test doesn't seem to be executed on the chipmunk, so will wait
for its results.

I will further look into it tomorrow morning.

[1] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=gharial&dt=2018-10-02%2006%3A30%3A47
[2] - https://www.csee.umbc.edu/portal/help/architecture/aig.pdf

-- 
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com