heap corruption in DBD::Oracle

I have a perl script which updates some (large) tables from text-files,
inserting, updating and deleting records. For performance reasons the
inserts, updates and deletes are queued together and then done with an
array operation when there are "enough". Now I've added another table
and the script dies with:

*** glibc detected *** double free or corruption (fasttop): 0x0000000001450ad0 ***

in ora_execute_array (called from line 854 in Oracle.pm)

The relevant part of the script looks like this:

sub insert {
my ($self, $inserts) = @_;
my $dal = $self->{_dal};

my $sth_insert_data = $dal->{_dbh}->prepare_cached("insert into data(id, real) values(?, ?)");
my $sth_insert_datacoord =
$dal->{_dbh}->prepare_cached(
"insert into datacoords(period_start, period_end, data_id, set_id)
values(?, ?, ?, ? )"
);

my $sth_insert_fact_comext =
$dal->{_dbh}->prepare_cached(
"insert into fact_comext(period_start, reporter, partner, product, data_id, real)
values(?, ?, ?, ?, ? ? )"
);

my @values = map { $_->{real} } @$inserts;
my $data_ids = $self->seq_numbers(scalar(@$inserts));
$sth_insert_data->execute_array({}, $data_ids, \@values); # <-- WORKS

unless ($time_period_id) {
my @tp = $dal->datasets({ path => [ 'time_period']});
die "time_period not found" if @tp < 1;
die "time_period not unique" if @tp > 1;
$time_period_id = $tp[0]->id;
}

my @ps = map { $_->{period}[0] } @$inserts;
my @pe = map { $_->{period}[1] } @$inserts;
$sth_insert_datacoord->execute_array({}, \@ps, \@pe, $data_ids, $time_period_id); # <-- WORKS

my @prod_ds = map { $_->{ds}->id } @$inserts;
$sth_insert_datacoord->execute_array({}, undef, undef, $data_ids, \@prod_ds); # <-- WORKS

my @rep_ds = map { $_->{reporter}->id } @$inserts;
$sth_insert_datacoord->execute_array({}, undef, undef, $data_ids, \@rep_ds); # <-- WORKS

my @part_ds = map { $_->{partner}->id } @$inserts;
$sth_insert_datacoord->execute_array({}, undef, undef, $data_ids, \@part_ds); # <-- WORKS

{
my @ps1 = @ps;
my @rep_ds1 = @rep_ds;
my @part_ds1 = @part_ds;
my @prod_ds1 = @prod_ds;
my @values1 = @values;

$sth_insert_fact_comext->execute_array({}, \@ps1, \@rep_ds1, \@part_ds1, \@prod_ds1, $data_ids, \@values1); # <--- CRASH
}

$dal->{_dbh}->commit();
$inserted += @$inserts;
}

There are six calls to execute_array in the routine. The first five
(marked "WORKS") work fine, the sixth crashes. The copying of the arrays
before the sixth execute_array was a feeble (and unsuccessful) attempt
at a workaround.

Has anybody run into this problem before or do I need to dig deeper?

Version information:

DBD::Oracle 1.19
DBI 1.58
perl v5.8.8 built for x86_64-linux-thread-multi
Red Hat Enterprise Linux AS release 4 (Nahant Update 5)
Oracle 10.2.0.1.0

hp

--
_ | Peter J. Holzer | If I wanted to be "academically correct",
|_|_) | Sysadmin WSR | I'd be programming in Java.
| | | hjp@xxxxxxxxx | I don't, and I'm not.
__/ | http://www.hjp.at/ | -- Jesse Erlbaum on dbi-users

Attachment: pgpiqeuqjORre.pgp
Description: PGP signature

Relevant Pages

  • Re: slow logon
    ... in the logon time have been reduced so far. ... update the time for the clients that talk to the dc as the logon server. ... start up script in AD to apply updates to the clients machines which is ...
    (microsoft.public.windows.server.active_directory)
  • Re: slow logon
    ... that i tell the software not to restart the services when new virus updates ... update the time for the clients that talk to the dc as the logon server. ... > We have a Start up script and a login script for users as well as> folder ...
    (microsoft.public.windows.server.active_directory)
  • Re: slow logon
    ... > we have symantec av 10 and yes the software looks for updates from a> local ... > i told the programmer that the domain controller at each site will> check ... > update the time for the clients that talk to the dc as the logon> server. ... >>> We have a Start up script and a login script for users as well as ...
    (microsoft.public.windows.server.active_directory)
  • Re: slow logon
    ... DNS is AD integrated and we have dhcp servers ... All clients are XP pro sp2. ... Not everyday is updates applied ... Make sure that the script detects if that update already has been made, ...
    (microsoft.public.windows.server.active_directory)
  • Re: slow logon
    ... that i tell the software not to restart the services when new virus updates ... i noticed that the login script created by the programmer has: ... update the time for the clients that talk to the dc as the logon server. ... start up script in AD to apply updates to the clients machines which is ...
    (microsoft.public.windows.server.active_directory)