Re: status of IO Channels on Windows?

[Rick Jones <rick jones2 hp com>]

Tor Lillqvist wrote:
> Rick Jones writes:
>  > I was wondering about the status of IO Channels under Windows -
>  > specificially with respect to SOCKETs (the capitalization is
>  > Windows', not mine :) I ask because:
>
>  > Currently full support is available on UNIX platforms, support for 
>  > Windows is only partially complete.
>
> There are some gotchas, but many things work without problems... Is
> that vague enough? ;)

Yep.  And with the little "learning program" I'm writing under linux, 
I've had fun with getting things non-buffered and NULL encoded :)  A bit 
of a mindwarp for me.  Probably good for me, but had me grumbling a 
couple times thinking "this isn't a blankety blank socket, it is a FILE 
stream!" :)

I take it that the deprecated status of g_io_channel_[read|write] is 
"for real" and not just to scare people away?

> One gotcha is that setting up a watch on a GIOChannel to a socket puts
> the socket in non-blocking mode. This doesn't happen on Unix. But
> then, if you are using watches, your program presumably should use
> non-blocking IO anyway (otherwise, why not just use blocking socket
> API calls directly), so this shouldn't be a big deal.

I suspect I can deal with that.

> There are also differences in how and when the callbacks are
> called. There are some open bugs on this (see for instance #331214)
> and things may change a bit (to make the behaviour more like on Unix)
> soon. Anyway, your best bet is to write your code so that when you get
> a callback for G_IO_IN, you keep looping reading as much as you can
> get until you get the G_IO_STATUS_AGAIN indication.

Ditto.

> (BTW, see http://www.iki.fi/tml/fosdem-2006.pdf for some hints.)

So, has it been compiled P64 yet?-) Don't yo just love how 64-bit 
windows went P64 rather than LP64 ?-)

rick jones