On Mon, Jul 07, 2008 at 08:24:54PM -0700, Archon810 wrote:
> So I've been tracing this extremely weird segfault in my code for
> about half a year now mostly due to the evasive nature of the crash
> and finally traced it down to these specific steps:
> - perl 5.10 (happened in 5.8 too)
> - DBI 1.605
> - DBD::mysql 4.007
> - $dbh = DBI->connect($dsn, $db_user, $db_passwd,
> {mysql_auto_reconnect => 1})
> - establish connection
> - kill the connection on the server side or let it time out
> - call $dbh->ping
> - experience a consistent seg fault
>
> Now, if I called DBI->connect_cached() instead of ping() and DBI-
> connect(), which is what I was doing before, the crash would occur as
> well.
>
> Here's the kicker: if I remove mysql_auto_reconnect, no crash occurs
> during $dbh->ping.
>
> Is there something in DBI that would do this? I'm doing additional
> analysis and talking to the writer of DBD::mysql in hopes of solving
> this problem but it looks like DBD::mysql doesn't implement its own
> ping(), so it's probably something DBI does...
DBD::mysql uses the mysql_ping function in the mysql client library.
Disabiling mysql_auto_reconnect is probably a good idea anyway as
auto reconnection is unsafe (ie you could loose a lock and not know).
You're more likely to get help is you can post a stack trace from the
core file.
Tim.
