Stipe Tolj wrote: > > Aarno Syvänen wrote: > > > Using wapbox only for pushing (fetching is done by operator's gateway) > > seems to > > cause an idle loop, one thread (poll thread) taking up to 40 % of CPU > > time. > > > > I hope that this does not happen when wapbox is used for fetching (I > > cannot test > > this myself). > > > > The backtrace of the idle loop follows (taken from gdb when the poll > > thread is > > locked into the loop): > > > > 0x401df2b0 in poll () from /lib/libc.so.6 > > (gdb) bt > > #0 0x401df2b0 in poll () from /lib/libc.so.6 > > #1 0x080869f3 in gwthread_pollfd (fd=35, events=1, timeout=-1) at > > gwlib/gwthread-pthread.c:582 > > #2 0x08083226 in conn_wait (conn=0x81fe570, seconds=-1) at > > gwlib/conn.c:724 > > #3 0x0804c188 in read_from_bearerbox () at gw/shared.c:106 > > #4 0x0804ab2d in main (argc=2, argv=0xbffffda4) at gw/wapbox.c:433 > > #5 0x4012f65f in __libc_start_main () from /lib/libc.so.6 > > > > Adding sleep does help, and does not cause problems, when when only push > > is used > > (then wapbox is waiting only for administrative commands). But this > > 'solution' is hardly elegant.
> Ok, ceriously, I hope my SSL server side implementation in > gwlib/conn.c does not break things here. > > We have currently similar problem regarding gwthead_pollfd() on the > Cygwin platform, where the bearerbox crashes if using WAP config > layout. > > I was not able to reclaim from where the error is coming from :(( > > Aarno, are you running a SSL-enabled (at least compiled in) bearerbox > an wapbox? If yes, could you try the same situation with a > --disable-ssl build kannel to see if it depends with the SSL routine > stuff? Thanks. I am running with --disable-ssl, so ssl code is not the problem. Aarno
