Re: Pilot Link 0.12

On Thu, 2005-11-03 at 15:39 -0500, JP Rosevear wrote:
> gnome-pilot/gnome-pilot-conduits cvs now contains patches to build
> against recent pilot-link 0.12 releases.
> Please test and let me know of any problems.


one problem I found is that the pilot-link pi_accept() API contract has
changed.  Since 0.12pre the accept() call now does some dup fiddling so
that it can return the accepted socket with the same file descriptor as
the listen socket it started with.  Behind the scenes the listen socket
has been closed.  I guess this prevents careless client code from fd
leaks, but it seems like the wrong approach to solving that problem.  I
complained on the pilot-link-dev list, but nobody seemed to care - the
change has probably been around too long to get removed at this stage.

The upshot of this is that the pi_close(listen_sd) at the end of
gpilotd.c:pilot_connect should be surrounded by a
#ifndef PILOT_LINK_0_12

Other than that, I seem to remember the CVS version worked okay (I'm not
certain, because I didn't do much testing before carrying on to make
further netsync-related changes).

There was also a bug in the backup_conduit code (failure to initialise
some variables) that caused a segfault for me when modifying the conduit
params with the capplet.  Patch:
Index: conduits/backup/backup_conduit.c
RCS file: /cvs/gnome/gnome-pilot/conduits/backup/backup_conduit.c,v
retrieving revision 1.49
diff -u -p -r1.49 backup_conduit.c
--- conduits/backup/backup_conduit.c	3 Nov 2005 20:17:16 -0000	1.49
+++ conduits/backup/backup_conduit.c	1 Jan 2006 19:21:09 -0000
@@ -169,9 +169,9 @@ static void 
 save_configuration(ConduitCfg *c) 
 	gchar *prefix;
-	gchar **exclude;
+	gchar **exclude = NULL;
 	GList *iterator;
-	guint i;
+	guint i = 0;



Matt Davey      Did you hear that there's a group of South American Indians
mcdavey mrao cam ac uk 	   that worship the number zero?  Is nothing sacred?

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]