Re: possible problem with DBD::mysql and new mysql clients

On Sun, 14 Nov 2004, Andreas Nolte wrote:

> Hi rudy,
> 
> hopefully this email does not qualify me as a total noob, but I just wanted to

nope. Segfaults always mean a bug (well unless you are getting a segfault
because you ran out of stack space. In which case, you might be able to think 
of
it as a !bug?).

> point at a possible problem with DBD::mysql and new mysql clients.
> 
> We are running a few sites with mysql and OpenInteract as a web application
> server. Since now MySQL 4.1.7 is marked productive, we thought we should test
> it out.
> 
> So well, I upgraded to DBI 1.45 and DBD::2.9004 after installing the new
> client and all tests ran successfully. But when I start the web server, apache
> keeps on segfaulting.
> 

I saw something similar on the Krang list, are you by chance using a 
mysql_default_group connect option somewere?  The person on Krang said that the 
segfaulting went away when read_default_group= was removed form the connect 
options, BTA when it comes to segfaults, any change to code can stop them, only 
to have them pop up somwere else.

> So you might say - this is an OpenInteract bug or whatever, but this is what I
> tested:
> 
> - downgraded DBI to 1.44 and 1.43 and recompiled -> same result. I did this,
> because there where rumors about DBI 1.45 bugs. 
> - upgraded DBD::mysql to newes devel version -> same result. 
> - downgraded MySQL client to 4.0.21 gcc -> everything works ok.
> 
> The strange thing is, that all of my test scripts w/o apache worked ok, as did
> some OpenInteract test scripts that go w/o apache.
> 

This is mod_perl, right?  My first guess based on your description is that this
sounds like there might be an interaction between apache and the latest MySQL
client library that is causing a SIGSEGV (I wonder if php is having any 
problems?).

> I `d be suppling more thorough input, if I knew how to approach this problem.
> Any hints ? The only thing I did not test ( yet ) is reading and storing BLOB
> data with the new client. OpenInteract uses Apache::Session to store session
> data - maybe this is what breaks the code.
> 
> So - if I got more, I`ll tell you. So far I just wanted to alert you, that
> there might be a problem - but maybe I am just doing s.th. wrong. If you got
> an idea, how I could generate some more useful input for you - please just
> tell me.

1. A DBI->trace(9) of the offending section of code.  

2. And then You could compile with -g and give me a backtrace from the core
file, also versions of apache and if you are using mod_perl the version of that.
And if you can, an isolated test script that will cause the segfault with a
minimum number of dependences (ideally just apache, DBD::mysql and mod_perl (if 
you are using mod_perl))



-- 
MySQL Perl Mailing List
For list archives: http://lists.mysql.com/perl
To unsubscribe:    http://lists.mysql.com/[email protected]