The Trunk: Network-fbs.140.mcz
Locked
 
|
Frank Shearar uploaded a new version of Network to project The Trunk:
http://source.squeak.org/trunk/Network-fbs.140.mcz ==================== Summary ==================== Name: Network-fbs.140 Author: fbs Time: 1 March 2013, 10:39:25.076 am UUID: 9f6d4989-84fc-4303-87ab-406d2296fd73 Ancestors: Network-fbs.139 If the user gives up on trying to resolve a name, fail the socket connect completely. http://bugs.squeak.org/view.php?id=4025. =============== Diff against Network-fbs.139 =============== Item was changed: ----- Method: SocketStream class>>openConnectionToHostNamed:port: (in category 'instance creation') ----- openConnectionToHostNamed: hostName port: portNumber | hostIP | hostIP := NetNameResolver addressForName: hostName timeout: 20. + hostIP ifNil: [NetworkError signal: ('Cannot resolve {1}.' format: {hostName})]. ^self openConnectionToHost: hostIP port: portNumber! |
Re: The Trunk: Network-fbs.140.mcz
|
|
I am not sure how to reproduce the original problem (on Linux), but I'm
not sure if the proposed fix will work if the IPv6 networking is disabled in preferences. The resolver behaves differently depending on whether the new network primitives are being used, and may answer #[0 0 0 0] rather than nil. NetNameResolver useOldNetwork: true. NetNameResolver addressForName: 'foo' timeout: 20 ==> #[0 0 0 0] NetNameResolver useOldNetwork: false. NetNameResolver addressForName: 'foo' timeout: 20 ==> nil Dave On Fri, Mar 01, 2013 at 10:39:44AM +0000, [hidden email] wrote: > Frank Shearar uploaded a new version of Network to project The Trunk: > http://source.squeak.org/trunk/Network-fbs.140.mcz > > ==================== Summary ==================== > > Name: Network-fbs.140 > Author: fbs > Time: 1 March 2013, 10:39:25.076 am > UUID: 9f6d4989-84fc-4303-87ab-406d2296fd73 > Ancestors: Network-fbs.139 > > If the user gives up on trying to resolve a name, fail the socket connect completely. > > http://bugs.squeak.org/view.php?id=4025. > > =============== Diff against Network-fbs.139 =============== > > Item was changed: > ----- Method: SocketStream class>>openConnectionToHostNamed:port: (in category 'instance creation') ----- > openConnectionToHostNamed: hostName port: portNumber > | hostIP | > hostIP := NetNameResolver addressForName: hostName timeout: 20. > + hostIP ifNil: [NetworkError signal: ('Cannot resolve {1}.' format: {hostName})]. > ^self openConnectionToHost: hostIP port: portNumber! |
Re: The Trunk: Network-fbs.140.mcz
|
|
I reproduce it typically by not having a network connection!
OK, so the solution is more complicated: NetNameResolver >> #addressForName:timeout: must have a way of signalling not only a name lookup failure (which loops via the exception's defaultAction) but also a name lookup failure where the user gives up. Given the return values, how about result := NetNameResolver addressForName: 'foo' timeout: 20. (NetNameResolver useOldNetwork and: [result = #[0 0 0 0]]) ifTrue: ["Signal failure"]. (NetNameResolver useOldNetwork not and: [result isNil]) ifTrue: ["Signal failure"]. ? frank On 2 March 2013 16:02, David T. Lewis <[hidden email]> wrote: > I am not sure how to reproduce the original problem (on Linux), but I'm > not sure if the proposed fix will work if the IPv6 networking is disabled > in preferences. The resolver behaves differently depending on whether the > new network primitives are being used, and may answer #[0 0 0 0] rather > than nil. > > NetNameResolver useOldNetwork: true. > NetNameResolver addressForName: 'foo' timeout: 20 ==> #[0 0 0 0] > > NetNameResolver useOldNetwork: false. > NetNameResolver addressForName: 'foo' timeout: 20 ==> nil > > Dave > > On Fri, Mar 01, 2013 at 10:39:44AM +0000, [hidden email] wrote: >> Frank Shearar uploaded a new version of Network to project The Trunk: >> http://source.squeak.org/trunk/Network-fbs.140.mcz >> >> ==================== Summary ==================== >> >> Name: Network-fbs.140 >> Author: fbs >> Time: 1 March 2013, 10:39:25.076 am >> UUID: 9f6d4989-84fc-4303-87ab-406d2296fd73 >> Ancestors: Network-fbs.139 >> >> If the user gives up on trying to resolve a name, fail the socket connect completely. >> >> http://bugs.squeak.org/view.php?id=4025. >> >> =============== Diff against Network-fbs.139 =============== >> >> Item was changed: >> ----- Method: SocketStream class>>openConnectionToHostNamed:port: (in category 'instance creation') ----- >> openConnectionToHostNamed: hostName port: portNumber >> | hostIP | >> hostIP := NetNameResolver addressForName: hostName timeout: 20. >> + hostIP ifNil: [NetworkError signal: ('Cannot resolve {1}.' format: {hostName})]. >> ^self openConnectionToHost: hostIP port: portNumber! > |
Re: The Trunk: Network-fbs.140.mcz
|
|
On Sat, Mar 02, 2013 at 06:40:10PM +0000, Frank Shearar wrote:
> I reproduce it typically by not having a network connection! > > OK, so the solution is more complicated: NetNameResolver >> > #addressForName:timeout: must have a way of signalling not only a name > lookup failure (which loops via the exception's defaultAction) but > also a name lookup failure where the user gives up. Given the return > values, how about > > result := NetNameResolver addressForName: 'foo' timeout: 20. > (NetNameResolver useOldNetwork and: [result = #[0 0 0 0]]) > ifTrue: ["Signal failure"]. > (NetNameResolver useOldNetwork not and: [result isNil]) > ifTrue: ["Signal failure"]. > > ? > > frank > I would think that just a check for nil or #[0 0 0 0] would be sufficient, regardless of the useOldNetwork setting. (hostIP isNil or: [hostIP = #[0 0 0 0]]) ifTrue: ["Signal failure"]. I am also not sure if different platforms handle this the same way (I would want to at least check Windows, though I can't do that right now). But my guess is that checking for nil or #[0 0 0 0] is likely to do the right thing an any case, regardless of platform differences (if any). Dave > On 2 March 2013 16:02, David T. Lewis <[hidden email]> wrote: > > I am not sure how to reproduce the original problem (on Linux), but I'm > > not sure if the proposed fix will work if the IPv6 networking is disabled > > in preferences. The resolver behaves differently depending on whether the > > new network primitives are being used, and may answer #[0 0 0 0] rather > > than nil. > > > > NetNameResolver useOldNetwork: true. > > NetNameResolver addressForName: 'foo' timeout: 20 ==> #[0 0 0 0] > > > > NetNameResolver useOldNetwork: false. > > NetNameResolver addressForName: 'foo' timeout: 20 ==> nil > > > > Dave > > > > On Fri, Mar 01, 2013 at 10:39:44AM +0000, [hidden email] wrote: > >> Frank Shearar uploaded a new version of Network to project The Trunk: > >> http://source.squeak.org/trunk/Network-fbs.140.mcz > >> > >> ==================== Summary ==================== > >> > >> Name: Network-fbs.140 > >> Author: fbs > >> Time: 1 March 2013, 10:39:25.076 am > >> UUID: 9f6d4989-84fc-4303-87ab-406d2296fd73 > >> Ancestors: Network-fbs.139 > >> > >> If the user gives up on trying to resolve a name, fail the socket connect completely. > >> > >> http://bugs.squeak.org/view.php?id=4025. > >> > >> =============== Diff against Network-fbs.139 =============== > >> > >> Item was changed: > >> ----- Method: SocketStream class>>openConnectionToHostNamed:port: (in category 'instance creation') ----- > >> openConnectionToHostNamed: hostName port: portNumber > >> | hostIP | > >> hostIP := NetNameResolver addressForName: hostName timeout: 20. > >> + hostIP ifNil: [NetworkError signal: ('Cannot resolve {1}.' format: {hostName})]. > >> ^self openConnectionToHost: hostIP port: portNumber! > > |
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |