FastSocketStream (was: FTP ConnectionTimedOut on 3.9)

stéphane ducasse ducasse at iam.unibe.ch
Thu Sep 8 12:05:06 UTC 2005


Goran

we pushed your code in 3.9alpha
and renamed the class FastScoketStream into Socket, so can you have a  
look to see
and eventually publish your fix in the sqf source

Stef
On 8 sept. 05, at 01:12, goran at krampe.se wrote:

> Hi!
>
> Since this smelled like an FastSocketStream issue I looked into it  
> - and
> found the issue:
>
> Jecel Assumpcao Jr <jecel at merlintec.com> wrote:
>
>> Marcus Denker wrote on Wed, 7 Sep 2005 17:31:40 +0200
>>
>>>   -> FastSocketStream replaced SocketStream... you could file in
>>>        the SocketStream from 3.8 to see if this makes a difference.
>>>
>>
>> I started having problems in Celeste so decided to load  
>> FastSocketStream
>> into the 3.8 image I am typing this in, but things stopped working
>> entirely (probably the ConnectionTimedOut: error, but I am not  
>> sure). I
>>
>
> In fact, first I thought this could have to do with the fact that FSS
> (if you read the class comment) signals ConnectionClosed and
> ConnectionTimedOut exceptions, which can affect old client code (which
> should be fixed IMHO).
>
> A "hack" to get the old and faulty behavior - send "shouldSignal:  
> false"
> to the stream.
>
> But... nah, then I started looking at the complex looking upToAll:
> method (we need a new prim btw to make it better!) but eventually  
> found
> the much simpler cause in #sendCommand:. It doesn't flush. Oops.
>
> So change it to:
>
> sendCommand: aString
>     self nextPutAllFlush: aString, String crlf
>
> After that it seems to work (you can test using a filelist, click at
> UIUC on the bottom).
>
>
>> On the other hand, I loaded FastSocketStream into a 3.8 image with
>> Seaside and it was a huge improvement there and has worked  
>> perfectly for
>> a couple of months now.
>>
>
> Yeah, I actually have debugged it pretty thoroughly and use it myself.
> But evidently sendCommand: is only used by client code - not server
> code. :)
>
>
>> -- Jecel
>>
>
> regards, Göran
>
>




More information about the Squeak-dev mailing list