RE: problems with GIOChannel

[Pablo <pawelp mail softsystem pl>]

W liście z czw, 08-05-2003, godz. 10:33, martyn.2.russell@bt.com pisze: 
> No, to remove the watch, you return FALSE in the callback function, as
> explained here:
> http://developer.gnome.org/doc/API/2.0/glib/glib-IO-Channels.html#GIOFunc

I forgot about it. Does it mean that after I issue
g_io_channel_shutdown() function, my callback function will be notified
about it and I'll have a way to decide if I should return FALSE?

> I did have a similar problem, I found that I was being notified that there
> was input ready to read (via the g_io_channel_add_watch() callback) and was
> reading the input, and that was fine.  In some instances (during early the
> stages of development) I was trying to read MORE than was waiting for me on
> the socket.  This mean't that the read function was waiting for more input
> and the rest of the application stopped dealing with events until such time.

That's exactly my problem. Socket is waiting for input (^C in gdb says
it's in read function). I was just wondering why is the socket waiting
for more input, why does it allocate internal buffers when I pass my own
buffer to g_io_channel_read_chars. But it's not important right now
since GIOChannel's struct member "use_buffer" set to FALSE solves my
problem.

> Also, just out of interest, have you thought about using gnet instead?? - I
> moved over to it for cross-platform compatibility, it makes light work of
> sockets and is fairly straight forward to use:
> http://www.gnetlibrary.org/

No I haven't. I just took a look at the website. It seems to be based on
glib so using glib alone should do the same job with cross-platform
compatibility. But structers and functions used in gnet lib are
interesting and maybe I'll give it a try.

Thanks for a reply.
Pablo