[squeak-dev] 20s send timeout in Socket >> sendSomeData:startIndex:count:

Frank Shearar frank.shearar at angband.za.org
Fri Feb 18 11:00:35 UTC 2011


On 2011/02/18 08:32, Frank Shearar wrote:
> On 2011/02/18 03:46, Matthew Fulmer wrote:
>> We ran into an issue in cobalt, and finally tracked it down to
>> the hardcoded 20 second timeout in Socket>>
>> sendSomeData:startIndex:count: . We need to make the timeout
>> configurable, ideally turning it off. The read timeouts in
>> Socket and SocketStream are configurable, kind of. Has anybody
>> done the same for the write timeouts?
>
> Not I, but it's a bit glaring now you point it out: explicit timeouts
> for connecting, and reading, but not writing!
>
> How about hollowing out sendSomeData:startIndex:count: and making it
> call a new method sendSomeData:startIndex:count:for: that takes an
> explicit timeout? (sendData: and sendUDPData:toHost:port: also use
> hard-coded timeouts.) That leaves the rest of the system untouched while
> you get an explicit timeout.

I submitted a proposal in Network-fbs.108, in the Inbox.

frank



More information about the Squeak-dev mailing list