FROM : Tobias Peciva
DATE : Fri Jan 04 03:12:43 2008
Hey List,
We're seeing some interesting NSSocketPortNameServer behavior in 10.5
- so wondering if anyone else has come across this. There are two
issues...
First, customers report that on some machines, the following call
fails (i.e. returns NO):
[[NSSocketPortNameServer sharedInstance] registerPort:port
name:@"unique.name.goes.here"]
According to the dox, this would only happen if another port is
registered with the same name - but this does not appear to be the
case, because calling portForName: yields nothing, and
removePortForName: also fails. I have also confirmed that the port
passed in is non-nil and valid. Interestingly enough, the problem only
started appearing on Intel iMacs after one of the software updates in
mid-December. Everything works fine on PowerPC Macs and non-iMac Intel
machines.
Second, I may have misunderstood how NSSocketPortNameServer works, but
I created two test applications that create one socket port each
(different port numbers, obviously), then attempt to register these
ports with the same name. The NSPortNameServer dox suggest that the
second app should fail to register its port, because another port has
already been registered with the same name. But in actual fact, both
ports are successfully registered with the same name.
So, bugs or features? What am I missing? Any ideas?
Thanks muchly!
Peace,
Tobias
--
Tobias Peciva
Team Leader / Software Architect
Pharos Systems - http://www.pharos.com/
DATE : Fri Jan 04 03:12:43 2008
Hey List,
We're seeing some interesting NSSocketPortNameServer behavior in 10.5
- so wondering if anyone else has come across this. There are two
issues...
First, customers report that on some machines, the following call
fails (i.e. returns NO):
[[NSSocketPortNameServer sharedInstance] registerPort:port
name:@"unique.name.goes.here"]
According to the dox, this would only happen if another port is
registered with the same name - but this does not appear to be the
case, because calling portForName: yields nothing, and
removePortForName: also fails. I have also confirmed that the port
passed in is non-nil and valid. Interestingly enough, the problem only
started appearing on Intel iMacs after one of the software updates in
mid-December. Everything works fine on PowerPC Macs and non-iMac Intel
machines.
Second, I may have misunderstood how NSSocketPortNameServer works, but
I created two test applications that create one socket port each
(different port numbers, obviously), then attempt to register these
ports with the same name. The NSPortNameServer dox suggest that the
second app should fail to register its port, because another port has
already been registered with the same name. But in actual fact, both
ports are successfully registered with the same name.
So, bugs or features? What am I missing? Any ideas?
Thanks muchly!
Peace,
Tobias
--
Tobias Peciva
Team Leader / Software Architect
Pharos Systems - http://www.pharos.com/
| Related mails | Author | Date |
|---|---|---|
| Tobias Peciva | Jan 4, 03:12 | |
| Scott Stevenson | Jan 10, 08:35 |






Cocoa mail archive

