Closed GoogleCodeExporter closed 9 years ago
GoogleCodeExporter
commented
9 years ago Fine, but on the other side:
Why on heaven would someone use DBD::PgPP over DBD::Pg?Original comment by reini.urban on 11 Jul 2014 at 5:49
GoogleCodeExporter
commented
9 years ago Because DBD::Pg segfaults on perl 5.6Original comment by todd.e.rinaldo on 11 Jul 2014 at 6:08
GoogleCodeExporter
commented
9 years ago In general: We are using a compiler, so we don't need to fallback to a pure
perl version. At all. This is nonsense. So low prio.
This compiler does not segfault when loading a XS.
However the same error appears with the XS version. It's clearly a DBI problem,
unrelated to IO::Socket
#! perl
use strict;
use DBI;
$ENV{DBI_TRACE} = 2;
my ($db, $user, $port) = ('rurban', 'rurban', 5433);
my $dbh = DBI->connect("dbi:Pg:dbname=$db;port=$port", $user, '',
{
'RaiseError' => 1,
'PrintError' => 1
}
);
vs
#! perl
use strict;
use DBI;
$ENV{DBI_TRACE} = 2;
my ($db, $user, $port) = ('rurban', 'rurban', 5433);
my $dbh =
DBI->connect("dbi:PgPP:dbname=$db;port=$port;path=/var/run/postgresql/", $user,
'',
{
'RaiseError' => 1,
'PrintError' => 1,
}
);
Original comment by reini.urban on 11 Jul 2014 at 6:45
GoogleCodeExporter
commented
9 years ago IO::Socket is outside the scope of the issue I raise here, so please ignore
that as a course of further investigation.Original comment by erin.schoenhals on 11 Jul 2014 at 7:04
GoogleCodeExporter
commented
9 years ago Erin: If it would have been a DBD:PgPP issue alone, it would have been related
to IO::Socket missing IO::Handle.Original comment by reini.urban on 12 Jul 2014 at 7:53
GoogleCodeExporter
commented
9 years ago Original comment by reini.urban on 14 Jul 2014 at 5:02
GoogleCodeExporter
commented
9 years ago It might be related either to wrong compile-time AMT method overload tables.
Add in DBI.pm:663
my $connect_closure = sub {
my ($old_dbh, $override_attr) = @_;
if ($ENV{DBI_TRACE} > 1) {
require Devel::Peek;
#warn "connect_closure: ".Devel::Peek::Dump([$attr,\%attributes, $override_attr]);
Devel::Peek::Dump($drh);
warn "drh: ".ref $drh, ", ISA: ",join(" ",@DBI::ISA);
}
and see the $drh AMT magic, and compare compiled vs uncompiled with -Do
pcc -O3 -v5 -A t/pg.pl -- -Do
vs
p -Do t/pg.pl
Or to DBI using it's own private magic extension ~ (PERL_MAGIC_ext)
=>
SV = IV(0x2294ef0) at 0x2294f00
REFCNT = 2
FLAGS = (PADMY,ROK)
RV = 0x226e810
SV = PVHV(0x23ff420) at 0x226e810
REFCNT = 3
FLAGS = (OBJECT,RMG,SHAREKEYS)
MAGIC = 0x22770b0
MG_VIRTUAL = &PL_vtbl_pack
MG_TYPE = PERL_MAGIC_tied(P)
MG_FLAGS = 0x02
REFCOUNTED
MG_OBJ = 0x22e1218
SV = IV(0x22e1208) at 0x22e1218
REFCNT = 1
FLAGS = (PADMY,ROK)
RV = 0x226eca8
SV = PVHV(0x23ff400) at 0x226eca8
REFCNT = 1
FLAGS = (OBJECT,RMG,SHAREKEYS)
MAGIC = 0x220af30
MG_VIRTUAL = 0
MG_TYPE = PERL_MAGIC_ext(~)
MG_FLAGS = 0x02
REFCOUNTED
MG_OBJ = 0x226e8b8
SV = PVMG(0x21f4640) at 0x226e8b8
REFCNT = 1
FLAGS = (OBJECT)
IV = 0
NV = 0
PV = 0x217bfb0 "\25\2\20\0\0\0\0\0\1\0\0\0\0\0\0\0\20\350&\2\0\0\0\0\20\335\241\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0@_8\2\0\0\0\0\20\335\241\0\0\0\0\0\0\0\0\0\0\0\0\0\222\6\0\0\0\0\0\0p`$\2\0\0\0\0\10\355&\2\0\0\0\0\300\354&\2\0\0\0\0P\355&\2\0\0\0\0h\355&\2\0\0\0\0\0\0\0\0\0\0\0\0P\0\0\0\0\0\0\0 \355&\2\0\0\0\0\0\0\0\0\0\0\0\0"\0
CUR = 152
LEN = 160
STASH = 0x2282d78 "DBD::Pg::dr_mem"
MG_PTR = 0x217bfb0 ""
STASH = 0x22830c0 "DBI::dr"
ARRAY = 0x220b350 (0:9, 1:6, 3:1)
hash quality = 90.0%
KEYS = 9
FILL = 7
MAX = 15
RITER = -1
EITER = 0x0
STASH = 0x22830c0 "DBI::dr"
ARRAY = 0x0
KEYS = 0
FILL = 0
MAX = 7
RITER = -1
EITER = 0x0
Original comment by reini.urban on 15 Jul 2014 at 10:25
GoogleCodeExporter
commented
9 years ago Work in branch dbi_ima:
- Save XSUBANY.any_long for all XSUB's, but some use any_ptr,
pointing to the ~ mg_ptr dbi_ima.
- This DBI_MAGIC ~ holds the dbi_ima_t struct, which stores some ptrs: stash,
gv, and for ithreads and BROKEN_DUP_ANY_PTR my_perl also.
The stash and gv is needed to install and find DBI methods. (A horrible
implementation. In a normal world I would expect to set the ISA correctly or at
least use overload or tie magic to find the run-time added methods for the used
driver.)
Original comment by reini.urban on 17 Jul 2014 at 2:35
GoogleCodeExporter
commented
9 years ago Original comment by reini.urban on 17 Jul 2014 at 2:36
GoogleCodeExporter
commented
9 years ago The DBI_MAGIC ~ does not only store the dbi_ima_t struct, it also stores all
internal dbi structs, which are imp_drh_t (driver), imp_dbh_t (database),
imp_sth_t (statement), imp_fdh_t (field descriptor), imp_xxh_t (any).
you get the type from the stash name with "_mem" suffix, e.g. DBI::Pg::dr_mem
is a imp_drh_t.
Certain parts of these structs need to be specialized at init2, e.g. the ptrs
to the gv or stash, ...Original comment by reini.urban on 17 Jul 2014 at 3:24
GoogleCodeExporter
commented
9 years ago The best fix is this patch for DBI.pm:
iff --git DBI.pm DBI.pm
index 2f84796..9efc7c5 100644
--- DBI.pm
+++ DBI.pm
@@ -506,7 +506,8 @@ my $keeperr = { O=>0x0004 };
},
);
-while ( my ($class, $meths) = each %DBI::DBI_methods ) {
+INIT {
+ while ( my ($class, $meths) = each %DBI::DBI_methods ) {
my $ima_trace = 0+($ENV{DBI_IMA_TRACE}||0);
while ( my ($method, $info) = each %$meths ) {
my $fullmeth = "DBI::${class}::$method";
@@ -518,6 +519,7 @@ while ( my ($class, $meths) = each %DBI::DBI_methods ) {
}
DBI->_install_method($fullmeth, 'DBI.pm', $info);
}
+ }
}
Original comment by reini.urban on 18 Jul 2014 at 3:00
GoogleCodeExporter
commented
9 years ago This patch looks fine except it throws nasty runtime errors if DBI is loaded
after INIT. Any way we could change the patch to work in both cases?Original comment by todd.e.rinaldo on 18 Jul 2014 at 3:31
GoogleCodeExporter
commented
9 years ago Yes, this patch is just a dirty hack, and does not pass the DBI testsuite, and
does not work in all scenarios.
I'm working in a perlcc specific solution to detect DBI and add the missing
post-dl_init initializations by myself. The driver (e.g. DBD::Pg) can by
compiled in or not. Usually not, but relying on that is too fragile.Original comment by reini.urban on 18 Jul 2014 at 3:25
GoogleCodeExporter
commented
9 years ago Testing now the fix commit ef491cbcca76998124f6315c16a6930e9a1fbf84
Author: Reini Urban <rurban@cpanel.net>
Date: Thu Jul 17 12:41:59 2014 -0500
C 1.49_01: fix DBI #359
special case DBI_MAGIC methods,
add an extra call to DBI->_install_method to re-initialize compile-time
internal ima pointers. ('~' DBI_MAGIC)
also store XSUBANY.any_long and use a heuristic to skip storing
compile-time pointers. storing ptrs broke Tie::Hash::NamedCapture (90)
also add example t/pg.pl: DBI_TRACE=0 t/pg
Original comment by reini.urban on 18 Jul 2014 at 8:25
GoogleCodeExporter
commented
9 years ago here is a DBI patch that would avoid warnings
diff --git a/modules/DBI/DBI/DBI.pm b/modules/DBI/DBI/DBI.pm
index 2e9f6fe..92b8b9e 100644
--- a/modules/DBI/DBI/DBI.pm
+++ b/modules/DBI/DBI/DBI.pm
@@ -499,6 +499,8 @@ my $keeperr = { O=>0x0004 };
},
);
+# wrap the code in a sub
+sub init_install_method {
while ( my ($class, $meths) = each %DBI::DBI_methods ) {
my $ima_trace = 0+($ENV{DBI_IMA_TRACE}||0);
while ( my ($method, $info) = each %$meths ) {
@@ -512,6 +514,16 @@ while ( my ($class, $meths) = each %DBI::DBI_methods ) {
DBI->_install_method($fullmeth, 'DBI.pm', $info);
}
}
+} # end sub init_install_method
+# install methods at INIT time when compiled
+# view https://code.google.com/p/perl-compiler/issues/detail?id=359
+if ( !${^GLOBAL_PHASE} || ${^GLOBAL_PHASE} eq 'START' ) {
+ eval q/INIT { init_install_method() }/;
+} else {
+ init_install_method();
+}
+
+
{
package DBI::common;Original comment by nicolas....@gmail.com on 22 Jul 2014 at 4:58
GoogleCodeExporter
commented
9 years ago note that such a patch cannot be submitted to DBI
as you will not be able to establish a db connection during a begin block...
which in most of the case will be a non sense... except if you want to read
some values from a DB during the compilation in order to use them laterOriginal comment by nicolas....@gmail.com on 22 Jul 2014 at 5:28
Original issue reported on code.google.com by
erin.schoenhalson 11 Jul 2014 at 4:29Attachments: