On Fri, 2009-03-06 at 16:48 -0500, Dan Williams wrote:
> On Fri, 2009-03-06 at 13:37 -0800, Inaky Perez-Gonzalez wrote:
> > On Friday 06 March 2009, Dan Williams wrote:
> > > On Fri, 2009-03-06 at 12:52 -0800, Inaky Perez-Gonzalez wrote:
> > > > On Friday 06 March 2009, The Hereweb wrote:
> > > > > My card was activited on XOHM under Windows first.
> > > > >
> > > > > In both XOHM and Clearwire, each uses a hotline website for the user 
> > > > > to
> > > > > set up their billing account/cc#/etc.  XOHM is different however in
> > > > > that it uses OMA-DM as part of the device activation process whereas
> > > > > Clear does not currently.
> > > >
> > > > Yeah, I was able to activate in XOHM some time ago without OMA-DM, but I
> > > > am concerned that they might have changed the process--that's why I keep
> > > > looking for guinea pigs :)
> > >
> > > I tried the other day, but got some segfaults in wimaxd releated to
> > > pthreads.  This was rawhide though, and I didn't have enough time to
> > > install the glibc-debuginfo, but will do that soon.
> > 
> > Huh, that should not be happening (tm).
> 
> Yeah, I'll track that down, at least get a backtrace with symbols.

http://bigw.org/~dan/wimaxd-crash.txt

l5_sockets_dispatcher_Stop() is calling wmx_timedjoin_thread() with a
NULL threadid.  That's because l5_sockets_dispatcher_Start() failed to
initialize the supplicant, because the binary supplicant library is only
compiled against libcrypto.so.0.9.8, while Fedora Rawhide ships a newer
libcrypto.so.0.9.8j which is a more current version of OpenSSL than the
binary supplicant has been compiled against.  Any chance we can get
either (a) a rebuilt binary supplicant library, or even better (b) an
open-source implementation not based off proprietary Devicescape code so
you guys wouldn't have to keep rebuilding it every time OpenSSL releases
an update?

l5_sockets_dispatcher_Stop() probably shouldn't be trying to join with
g_hRequestHandlingThread unless that thread has actually been created,
which in this case it hasn't...

This works for me in L5SocketsDispatcher.c:

+ if (g_hRequestHandlingThread) {
        wmx_set_event(g_hStopRequestHandlingThreadEvent);
        if ( wmx_timedjoin_thread(g_hRequestHandlingThread, SHUTDOWN_TIMEOUT, 
NULL) != 0 )
                wmx_kill_thread(g_hRequestHandlingThread);
+ }

and allows wimaxd to gracefully exit when it can't load the supplicant
library.

Dan


_______________________________________________
wimax mailing list
[email protected]
http://www.linuxwimax.org/mailman/listinfo/wimax

Reply via email to