linux 64 bit

This error is not repeatable on demand. The 'how to repeat' is unknown.

out2.out The attached file ( out2.out ) gives more details about the
value of *net and *net->vio.

Noting that we have a broken network, the code seems to be trying to close the connection to the mysql db. Somewhere before we got here , net->buff was set to zero.
Then by following the backtrace, we see that net->write_pos will be set to net->buff, which is zero.
Since net->write_pos is zero, we core dump.

So, why is the program trying to write a command onto a connection is that broken.
Even though the net->vio is non-zero, the structure pointed to by net->vio is all zeros ( as verified by the out2.out file ). So, it seems that a check for ( net->vio != NULL ) is not good enough.

anybody out there ?

We fixed several things in mysql_close, close_slow_part and close_options - can you try to repeat the issue with latest 2.3 release please?
Can you also provide options like reconnect (p *mysql in gdb would be enough).

Thanks!

sorry - I do not have the ability to experiment.
This error was found on an older version of our product.
The most current version of our product already is using version 2.3 of the connector.

Besides, the files that I supplied in this ticket from MariaDb-connector v 2.1.x
are virtually identical to the 2.3 versions.

??

the content of the variable *mysql is already provided at line 393 of file rob.out
which is contained in the 'tarball.tgz'

Sorry, I meant mysql->net.vio - it looks like it is overwritten somehow.

the display of *net->vio is found at line 50 of the out2.out file that is attached to this ticket.

Some observations:

1) vio->type is VIO_CLOSED - value was assigned in function vio_close()

  vio->type= VIO_CLOSED;

  vio->sd=   -1;

but p *net->vio shows another value for sd:

sd = 54158592

2) net->buff was set to 0 in function net_end().
net_end is called inside function end_server(), before net->vio is set to 0.

  if (mysql->net.vio != 0)

  {

    init_sigpipe_variables

    set_sigpipe(mysql);

    vio_delete(mysql->net.vio);

    reset_sigpipe(mysql);

    mysql->net.vio= 0;    /* Marker */

  }

  net_end(&mysql->net);

p *net shows another value for vio:

vio = 0x33a6500

The value for vio is exactly the same as for vio->sd: 54158592 is 0x33a6500 - so it looks like an invalid write happens somewhere: either in MariaDB Connector/C or in your application.

With latest C/C 2.3 I wasn't able to reproduce the crash nor did valgrind report any invalid memory operations.

the key piece of information in this ticket is that there was a network failure.
Thus, error logic/handling is play a role in this situation.

my question from Dec 8...
why is the program trying to write a command onto a connection is that broken?
Even though the net->vio is non-zero, the structure pointed to by net->vio is all zeros ( as verified by the out2.out file ). So, it seems that a check for ( net->vio != NULL ) is not good enough.