dswifi problem? {accept&close)x7 = guru meditation

[matsuri]

I'm developing some sort of server and spent some time debugging a problem where it crashes when it accepts the 7th connection.

After many experiments, I decided to write a very simple test case, because I suspected a bug in the sdk somewhere... and the test case crashes, exactly at the 7th time!

So it probably is an SDK bug (my SDK was installed 2 days ago, got latest versions of everything).

I use an EzFlash Vi on a DSi.

Test case does just the following:

Code: Select all

void tcpserv(){
        int sd, psd,addrlen;
        struct sockaddr_in name,addr;
        sd = socket (AF_INET,SOCK_STREAM,0);
        name.sin_family = AF_INET;
        name.sin_addr.s_addr = htonl(INADDR_ANY);
        name.sin_port = htons(80);
        bind(sd, (struct sockaddr *)&name, sizeof(name));
        listen(sd,1);
        addrlen=sizeof(addr);
        while(1){
                do{
                        psd = accept(sd,(struct sockaddr *)&addr, &addrlen);
                } while(!psd);
                printf("%d,",psd);
                shutdown(psd,SHUT_RDWR);
                close(psd);
        }
}

The whole source to test is available at: http://pty.yi.org/dstcplistener_clean.c
Prebuild NDS at: http://pty.yi.org/dstcplistener.nds

[elhobbs]

have you tried this with a card other than the ezflash Vi? I have the Vi and the V. I have some issues when using the Vi sometimes - on things where the V seams to work ok.

[matsuri]

have you tried this with a card other than the ezflash Vi? I have the Vi and the V. I have some issues when using the Vi sometimes - on things where the V seams to work ok.

I would. I don't have more cards to try with, however.

Thankfully if somebody tries the test case I wrote, if it is reproduceable (or not, that would be interesting!) can be determined quite quick.

My gut tells me everybody who tries it will experience the have the same issue, however.

[Alkiera]

Downloaded your .nds to my hardware; DSLite with Datel GnM; same issue. Tho technically, it never gets to the 'close' portion of the code the 7th time, the connection stays open; had to manually close telnet that 7th time.

Alkiera

[MrAndini]

I have found that if you do shutdown followed immediately by close, the other end of the TCP link doesn't always see the close of the TCP connection, so it thinks the connection is still active.

In my code I put a delay of 60 frames in between shutdown and close and the other end always correctly sees the shutdown of the link.

I don't know if this is related, but may be worth you giving it a try, might help identify the problem at least.

[Quirky]

Try with closesocket() instead of close().

EDIT: hmm, maybe not as that changes the semantics of the example...