Posted by Jochem Maas on 10/11/05 15:08

Chuck wrote:
> Yes, prefork is what I am using.

ah ok.

> Also made sure and linked with pthread

php + threads are not bedfellows - maybe try dropping
pthread as your AcceptMutex apache directive?

otherwise it could be time to crack open gdb and run apache
through it try and figure out exactly where the problem lies:

something like (I googled for it):

$> gdb /usr/local/apache/bin/httpd
> run -X -DSSL -f /usr/local/apache/conf/httpd.conf
<then view a pagfe and grab a BT>

also/alternatively try recompiling php with only the --with-apxs configure
flag - if that works recompile, each time adding a single flag,
this may pinpoint the problem some more.

.... and you could try stepping down to 5.0.4 and wait until 5.1 has
a stable release.

> as well.
>
> -CC
>
> On 10/10/05, *Jochem Maas* <jochem@iamjochem.com
> <mailto:jochem@iamjochem.com>> wrote:
>
> Chuck wrote:
> > After 3 months of headache and intermittent seg faults on Solaris
> 9, we
> > decided to give SLES 9 a try.
> >
> > Now we get continuous seg faults.
> >
> > I am running apache 2.0.54 on SLES 9 with all updates, and Oracle
> > 10.2.0.1 <http://10.2.0.1><http://10.2.0.1>.
> > (everything is the 32-bit flavor)
> >
> > I built PHP as follows:
> > adcinfops02:/usr/local/httpd-2.0.54/htdocs #./configure
> > --prefix=/usr/local/php-5.0.5
> --with-apxs2=/usr/local/httpd-2.0.54/bin/apxs
>
> long shot: which worker module do you have enabled in httpd.conf?
> you can only use 'pre-fork' with php.
>
> > --enable-cli --enable-debug
> --with-config-file-path=/usr/local/php-5.0.5/lib
> > --with-zlib --with-bz2 --enable-ftp --with-gettext
> > --enable-mbstring --with-ncurses
> --with-oci8=/u01/app/oracle/product/10.2
> > --enable-session --enable-sockets --enable-shared --disable-xml
> > --disable-libxml --disable-dom --with-jpeg-dir=/usr/lib
> > --with-png-dir=/usr/lib --with-zlib-dir=/usr/lib
> --with-xpm-dir=/usr/lib
> > --disable-simplexml --without-pear
> >
> > Here is the simply php page that seg faults _every_ time it is
> called. (On
> > solaris, it only seg faulted 25% of the time)
> > <?php
> >
> > putenv("ORACLE_HOME=/u01/app/oracle/product/10.2");
> > putenv("ORACLE_SID=ADCDM02");
> > putenv("TNS_ADMIN=/var/opt/oracle");
> >
> >
> > $ora = ociplogon("dm","mypassword", "MYSID");
> >
> > $stmt=OCIParse($ora, "select USERNAME from dba_users");
> > $res=OCIExecute($stmt);
> >
> > $rows = OCIFetchstatement($stmt, $results);
> >
> > print "DEBUG: rows=$rows<BR>\n";
> >
> > $keys = array_keys($results);
> > foreach($keys as $key)
> > {
> > print "$key<BR>\n";
> > }
> > for($i=0; $i < $rows; $i++)
> > {
> > print " " . $results["USERNAME"][$i] . "<BR>\n";
> > }
> >
> > OCIFreeStatement($stmt);
> > OCILogoff($ora);
> >
> > ?>
> >
> >
> > I can run this script 100 times from the command line ( # php
> oratest.php )
> > and not a single seg fault.
> >
> > Anyone have any idea why this is happenning?
> >
> > Also, is there a config, any OS, any version, and combination AT
> ALL, where
> > PHP is reliable when using Oracle 10g as the back end? Any
> configuration at
> > all?
> >
> > Thx,
> > CC
> > --
> >
>
>
>
>
> --
>
> To reply to this message, remove 'N_O_S_P_A_M' from the reply-to address.

[Back to original message]