Re: [Pan-users] 0.14 - attempted fix of the "freeze when connection is lost

[Alberto BARSELLA]

Hi,

On Tue, 13 May 2003 at 09:34:01AM -0400, Rinaldi J. Montessi wrote:

> Recompiled with your patch and tested it on some heavy downloads.  While 
> the sockets do wake up after going to sleep (the intended purpose) I 
> have had a couple of segfaults.  No data corruption, however.

I'd have expected more corruption than segfaults....
I'm having a look at what I've changed.  I need to check the glib
documentation about a couple of functions to see if one of the calls I
added can be the source of the segfault.

> I do get the occasional "Error reading from socket" message in my log.

Well, this is not surprising.  When a timeout is encountered, I set the
error flag for the socket to have pan drop the connection and clean up
its data structures.  It would be possible to provide a nicer message,
but that may require some more in-depth modification.  I'll let the
developers do it if people feel it's needed.

> Any pertinent data I can provide?

The only place I can see where I may have added a segfault is when a new
connection is created, but unless there's a way to have pan keep its log
onto the disk it won't be easy to determine if it's that.

I'll try to fix stuff and send in a new patch shortly, then you can see
if it fixed the segfault.

Bye,
Alberto
--
Alberto BARSELLA
** Beliefs are dangerous. Beliefs allow the mind to stop functioning.
A non-functioning mind is clinically dead.  Believe in nothing... **