Adrian It might. IIRC, there is at least one IPV6-specific primitive method that is called w/o checking #useOldNetwork. My working image contains attempts at fixing that, so I should probably try a fresh image on the offending machines. Is there really a useOldNetwork setting? I do not see anything in 10505 dev. Bill -----Original Message----- From: pharo-project-bounces@lists.gforge.inria.fr [mailto:pharo-project-bounces@lists.gforge.inria.fr] On Behalf Of Adrian Lienhard Sent: Wednesday, January 13, 2010 4:27 AM To: Pharo-project@lists.gforge.inria.fr Subject: Re: [Pharo-project] Network tests - Pharo vs. Squeak Hi Bill, If I remember correctly the consensus was to enable the useOldNetwork setting (at least in 1.0 and until the new implementation works reliably). Somehow we forgot to do that. Wouldn't that solve the problem? Cheers, Adrian On Jan 12, 2010, at 23:27 , Schwab,Wilhelm K wrote:
Hello all,
I was recently reminded of network problems, such as a need to force the use of the "old network" on at least some Linux machines, and that I have had mixed results on Windows (not sure what is causing grief there, but suspects include 2k3 vs. xp, bad network setup in general and ipv6 in particular).
So far, I have only one useful method that establishes a connection queue, then uses SocketStream to connect to is as locahost, enters a loop on #getConnectionOrNil, and then finally does some I/O. As I said, results are mixed.
Another test that needs to be done is split the client and server across a real network. I can do that (perhaps subject to weird constraints) between win2k, 2k3 and Linux. I tend to think of Linux as the client, but I might be able to flip that around.
My expectation (I can't believe I'm typing this<g>) is that Squeak will perform better than Pharo, only because of the "new network" changes, which I believe to be the cause of the problems.
I envision something like a friendly root kit: get each machine to download both Pharo and Squeak, and install a small .mcz into each image. The idea is to allow many of us to try this on different networks and platforms, and to merge the results.
For remote testing, there would probably need to be a server class with start/stop methods, and the client side test methods would hopefully log what they can. On successful remote conversation, the client should be able to automatically log details about the server, having read the details over the network.
Suggestions on how to build this would be greatly appreciated. Can it all be done with Sunit, or is that asking too much?
Bill
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project