Will the real sqUnixSocket.c step forward?

John M McIntosh johnmci at smalltalkconsulting.com
Sat May 4 01:09:04 UTC 2002


>At 6:26 PM +0200 5/3/02, Ian Piumarta wrote:
>>Lex,
>>
>>I just rewrote aio and various bits of socket code.  Goran has the
>>new sources and when I hear it works for him I'll post them.
>>
>
>Well I've a copy too I'm trying to get to work on a freebsd 
>4.3-Release-p28 i386 box.
>
>But still have the connection abort problem
>

Ok found out that Ian shipped the code with #if 1 in the accept 
handler to Unconnect the listener socket on an accept failure. This 
isn't the usual logic, usually we ignore the accept failure (client 
closed the connection) and continue.

But this exposed an issue in the Comanche ConnectionHandler>>pvtnewListenLoop

It seems the code there on the accept failure just checks to see if 
the listener Socket>>isValid which it is, but of course now in 
Unconnected status, versus WaitingForConnection or some other valid 
state, so then this while loop runs really fast and locks up the 
image.

That's a bug. And I'll let Ian decide what to do with the aborting of 
the listen socket since flipping it to unconnected (does it get 
closed? or is it still actually listening) doesn't really work as 
intended...

Someone could review the Comanche code and consider what if... Or is 
that fixed now? Yes I'm working with an old version...
-- 
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
===========================================================================



More information about the Squeak-dev mailing list