Ned Konz wrote:
> 
> Rob Withers wrote:
> 
> > sqSocketRemotePort(peername  -1,  AF_INET  2,  socketFamily  10273)
> >
> > It looks like the socket family is wrong for some reason.   I don't
> > think I can test this one further, as I've never tested the contents of
> > a sockstruct before...sorry.
> 
> I'm sure you know this, but remember that some of these things are other
> than "int" sized (like unsigned short, etc.). Make sure you cast them
> into ints or unsigned ints before passing them to fprintf.

well, now that you mention it, there is something funny about int
representation in c isn't there?  Could you imagine a machine where the
representation doesn't matter?  (sic ;)   I am still poking around a bit
and found dome other indicators to the other problems, which I'll put in
another note.  I'll try playing with this a little more...thanks!


> I've been bit before by this...
> 
> Also, there's no guarantee that these numbers (error codes, flag values, etc.)
> have predictable values between operating systems; there should probably
> be a lookup table on one side or another (that is: either don't return raw C
> return
> values, or be able to decode them on the Smalltalk side).

> Hmm... what about making a primitive that just returns a lookup table
> that can be made into a Dictionary? It could be returned in a Smalltalk-supplied
> buffer, as a series of:

I'm mixed about this.  It makes a lot of sense to put it with the other
platform dependent stuff, but why couldn't that include a bit of the
image as platform dependent?  Perhaps an image segment?  I would suggest
making the name a Symbol.

>         name (NUL terminated string), value (4-byte integer value)
> 
> pairs.

Rob
 
> --
> Ned Konz
> currently: Stanwood, WA
> email:     ned at bike-nomad.com
> homepage:  http://bike-nomad.com

-- 
--------------------------------------------------
Smalltalking by choice.  Isn't it nice to have one!