FreeLists / openbeosnetteam / [openbeosnetteam] Re: Now it seems this thing works! :)

["David Reid" <dreid@xxxxxxxxxxxx>]

Finally - an independent test!!!

> [/boot/home/development/cvs/openbeos/net_kit/source/apps/ifconfig]> ./
> ifconfig -a
> sis9000: flags=39<UP,RUNNING,MULTICAST,BROADCAST> mtu 1500
>         inet 192.168.157.1 netmask 0xffffff00 broadcast 192.168.157.255
> loop0: flags=119<UP,RUNNING,MULTICAST,LOOPBACK> mtu 16384
>
> BTW, wouldn't it be better to add something more standard as the
> interface name? making it be the name of the chipset being used (the
> driver being used, actually) makes things confusing. I like the ethN
> style from Linux.

Well, suggestions welcome, patches even more so :)

>
> The sis9000 interface was added using ifconfig (ifconfig sis9000
> 192.168.157.1 up netmask 255.255.255.0).
>
> [/boot/home/development/cvs/openbeos/net_kit/source/apps/route]> ./
> route show
> Routing tables
>
> Internet:
> Destination      Gateway            Flags
> default          connect!
> 192.168.157.1      UG
> connect!
> 192.168.157.0    connect!
> 192.168.157.1      U

The problem is the "connect!" which is debugging code and can be removed...

>
> It needs better formating for the output, but it works. :) Note that I
> already added a default route (route add default 192.168.157.1).

Formatting is OK after above is corrected...

>
> [/boot/home/development/cvs/openbeos/net_kit/source/apps]> ./
> select_test
> ===============
> = Select test =
> ===============
> Trying with timeval...
> select gave 1
> Trying without timeval (= NULL)
> select gave 1
> Test complete.
>
> I *REALLY* think select() should be provided in a separate library.
> This is because we will have select() support in OpenBFS too (actually,
> it already works) but the way it is now apps will have to link against
> libnet.so even if all they want to do is use select() on files, not on
> sockets. What do you guys think?

Well, there's nothing wrong with having a seperate library...

> Marcus, I know you read this list... How about updating dosfs so it
> supports select() on files? It is *REALLY* easy to do that and I can
> help you. Thsi would give us something to test select() on files even
> before OpenBFS is ready (Axel has been lazy lately ;) ).

If it's that easy then maybe you could fix the support we have? It doesn't
timeout as it should and we still don't actually wait on a socket being
ready... Look at the code for more (soselect in net_server/socket.c)

>
> [/boot/home/development/cvs/openbeos/net_kit/source/apps]> ./test1
> Thread 1, starting test...
> Thread 2, starting test...
> Thread 1:
>        sockets created :   201
>        test time       :    10 seconds
>        average         :    20 sockets/sec
> Thread 2:
>        sockets created :   202
>        test time       :    10 seconds
>        average         :    20 sockets/sec
> Test complete.
>
> This test uses 100% of my CPU (PIII 800). Is this expected?

Not really. Full debugging is on by the looks of the figures as well...

>
> [/boot/home/development/cvs/openbeos/net_kit/source/apps]> ./test2
> ===================================
> = Socket creation and bind() test =
> ===================================
> Thread 1, starting test...
> Thread 2, starting test...
> Error: Socket could not be boundto an ephemereal port
> Code: -2147454953
> Desc: Can't assign requested address
>
> Now, what was the problem here?

Doh! Not sure but it needs fixing!

>
> [/boot/home/development/cvs/openbeos/net_kit/source/apps]> ./test3
> =====================================
> = Simultaneous socket creation test =
> =====================================
> Thread 1, starting test
> Trying socket 0
> Trying socket 1
> Trying socket 2
> Trying socket 3
> Trying socket 4
> Trying socket 5
> Trying socket 6
> Trying socket 7
> Trying socket 8
> Trying socket 9
> Thread 1: completed creating 10 sockets...
> binding socket 0
> Error: Failed to bind!
> Code: -2147454953
> Desc: Can't assign requested address
>
> And another problem.

Looks like the same one...

>
> In a few words... David, you *ROCK*.

Thanks :)

david