Re: ActivePerl 5.8.4 w/ DBD::Sybase hangs on connect

[Henri Asseily]

I've been running DBD::Sybase with the following patch since version 
1.01, which could remotely have an impact for you. This is the 1.02 
patch, but should work fine in 1.04. FYI in DBD::Sybase version 1.04, 
the modified line is dbdimp.c:2987.

Here's the issue in syb_st_finish()  ( when $sth->finish is called):
finish() is called. If flush_finish, then while the statement is active 
and the connect isn't dead, fetch all the rows one after the other 
until the return value is null.
Now in the case where there's an error in the dbh, you risk an infinite 
loop because the dbh isn't dead, the statement is still active, and 
you're trying to fetch all the rows. On every fetch the db connect 
returns the same dbh error that isn't doing you any good and isn't 
triggering any of the exit statements.

This case can and does happen in some arcane setups such as when using 
openswitch which doesn't kill your dbh but invalidates it. All we need 
to do is put a check for a dbh error.

--- DBD-Sybase-1.02/dbdimp.c.orig       2003-10-28 13:21:27.000000000 
-0800
+++ DBD-Sybase-1.02/dbdimp.c    2003-10-28 13:25:26.000000000 -0800
@@ -2762,7 +2762,8 @@
     connection = imp_sth->connection ? imp_sth->connection :
        imp_dbh->connection;

-    if (imp_dbh->flushFinish) {
+    if (imp_dbh->flushFinish && !(SvOK(DBIc_ERR(imp_dbh)))) {
+       /* only flush if there is no error, otherwise we risk an 
infinite
loop */
        if(DBIS->debug >= 2)
            PerlIO_printf(DBILOGFP, "    syb_st_finish() -> 
flushing\n");
        while (DBIc_ACTIVE(imp_sth) && !imp_dbh->isDead) {

On Jul 2, 2004, at 10:56 PM, Michael Peppler wrote:
On Fri, 2004-07-02 at 22:06, David Good wrote:
On Fri, Jul 02, 2004 at 08:37:43PM +0200, Michael Peppler 
<[EMAIL PROTECTED]> wrote:
On Fri, 2004-07-02 at 20:19, David Good wrote:
I'm having trouble with DBD::Sybase 1.04 on ActivePerl 5.8.4 (build 
810)
on Windows.  It seems to hang when connecting.  Here's a sample 
script
that exercises the problem:

DESTROY     DISPATCH (DBI::st=HASH(0x1c1ca7c) rc1/1 @1 g0 ima4 
pid#2184) at c:/Perl/site/lib/DBD/Sybase.pm line 99 via foo line 
10
    -> DESTROY for DBD::Sybase::st (DBI::st=HASH(0x1c1ca7c)~INNER) 
thr#15d4a54
    syb_st_finish() -> ct_cancel(CS_CANCEL_ALL)
I don't know why, but the ct_cancel() call often causes problems - in
particular if the client and the server aren't on the same network 
(i.e.
firewalls, routers, etc.)

Try adding the syb_flush_finish => 1 attribute in the connect() call,
and read about this attribute in the DBD::Sybase docs.
Well, that fixed it, but the client and server are on the same 
network, so
I'm not sure what the deal is.  Oh well, as long as it works :-)
It might also be a bug in the ct_cancel() implementation on Windows for
this version of the SDK.
Michael
--
Michael Peppler                              Data Migrations, Inc.
[EMAIL PROTECTED]                       http://www.peppler.org/
Sybase T-SQL/OpenClient/OpenServer/C/Perl developer available for short
or long term contract positions - http://www.peppler.org/resume.html