Bug #55028 for DBD-Oracle: DBD INVALID_HANDLE: OCILobGetLength

On Wed Apr 14 16:02:00 2010, MJEVANS wrote: Show quoted text
I think: DBD::Oracle::db ora_lob_length failed: (DBD INVALID_HANDLE: OCILobGetLength) at empty_lob.pl line 26. is the key. I think OCILogGetLength is passed a NULL handle. Martin -- Martin J. Evans Wetherby, UK

On Fri Apr 23 12:16:31 2010, PYTHIAN wrote: Show quoted text
Thanks, I will look at it but I still don't understand. If I return a NULL lob from a procedure and Oracle says, it is null so I'm not going to give you a real lob locator, I'm just going to given you NULL then why doesn't DBD::Oracle check for NULL before attempting to use it resulting in the invalid handle. I don't see why the onus should be on Perl code outside of DBI/DBD::Oracle to check for this as I can create a Perl example that does not even know about lob locators. Martin -- Martin J. Evans Wetherby, UK

Its a funny Oracle thing Lob locater can be in one of two states initialized or not. a Lob 'field' can be null or contain a lob locater. That Lob loacter must be initialized even if it points to nothing I think in the case of your code it is returning an uninitialized locater. The empty_clob and empyt_blob SQL function will insert an empty initialized locater in a field hence the reason one will normally never get and uninitialized locater. Hope this helps On Fri Apr 23 12:41:23 2010, MJEVANS wrote: Show quoted text

On Fri Apr 23 12:58:28 2010, PYTHIAN wrote: Show quoted text
It does, thanks John. I'll try empty_clob. Martin -- Martin J. Evans Wetherby, UK Show quoted text

On Mon Apr 26 04:00:16 2010, MJEVANS wrote: Show quoted text
loacter Show quoted text
never Show quoted text

Sorry John, I've been really snowed under lately. The original example I provided does work if you use empty_clob. For anyone reading this here it is: use strict; use warnings; use DBI; use DBD::Oracle qw(:ora_types); use Data::Dumper; my $h = DBI->connect; $h->do(<<'EOT'); create procedure p_mje(plob OUT NOCOPY clob) AS BEGIN plob := empty_clob(); END; EOT my $s = $h->prepare(<<'EOT', {ora_auto_lob => 0, ora_check_sql => 0}); begin p_mje(?); end; EOT my $lob; $s->bind_param_inout(1, \$lob, 20, {ora_type => ORA_CLOB}); $s->execute; print Dumper([$lob]); my $len = $h->ora_lob_length($lob); END { $h->do(q{drop procedure p_mje}); } Thanks for looking in to this. Martin -- Martin J. Evans Wetherby, UK

Tested by user

On Fri May 28 13:16:29 2010, PYTHIAN wrote: Show quoted text
Sorry to come back to this so late but in actual fact things don't quite work as you'd expect. If you use empty_clob to set up an out param in a procedure and then try and retrieve it via DBD::Oracle you get a segmentation violation: use strict; use warnings; use DBI; use DBD::Oracle qw(:ora_types); use Data::Dumper; my $h = DBI->connect; $h->do(<<'EOT'); create procedure p_mje(plob OUT NOCOPY clob) AS BEGIN plob := empty_clob(); END; EOT my $s = $h->prepare(<<'EOT', {ora_auto_lob => 0, ora_check_sql => 0}); begin p_mje(?); end; EOT my $lob; $s->bind_param_inout(1, \$lob, 20, {ora_type => ORA_CLOB}); $s->execute; print Dumper([$lob]); my $len = $h->ora_lob_length($lob); print "$len\n"; # the following segfaults # seems you need to test $len is not 0 first my $lob = $h->ora_lob_read($lob, 1, $len); END { $h->do(q{drop procedure p_mje}); } Personally I do think it is a bug that you get a segfault but at the very minimum it should be mentioned in the docs that if you attempt to read a lob with a zero length you'll cause your program to segfault. Martin -- Martin J. Evans Wetherby, UK

Here is a stack trace: Program received signal SIGSEGV, Segmentation fault. XS_DBD__Oracle__db_ora_lob_read (my_perl=0x8188008, cv=0x82f2198) at Oracle.xs:421 421 bufp_len = SvLEN(dest_sv); /* XXX bytes not chars? (lab: yes) */ #0 XS_DBD__Oracle__db_ora_lob_read (my_perl=0x8188008, cv=0x82f2198) at Oracle.xs:421 #1 0x00744469 in XS_DBI_dispatch (my_perl=0x8188008, cv=0x81fd888) at DBI.xs:3424 #2 0x080b12c0 in Perl_pp_entersub () #3 0x080af688 in Perl_runops_standard () #4 0x080adbb2 in perl_run () #5 0x08063ffd in main () Martin -- Martin J. Evans Wetherby, UK

I tried reproducing with the following but it does not segfault for me: #!/usr/bin/perl use Inline C => <<'END_C'; void test(UV length) { STRLEN bufp_len; SV *dest_sv; void *bufp; printf("length is %u\n", length); dest_sv = sv_2mortal(newSV(length*4)); printf("dest_sv = %p\n", dest_sv); SvPOK_on(dest_sv); bufp_len = SvLEN(dest_sv); bufp = SvPVX(dest_sv); } END_C test(0); Including it just in case it helps. Martin -- Martin J. Evans Wetherby, UK

This is still the case with 1.28. Martin -- Martin J. Evans Wetherby, UK