karn@allegra.UUCP (Phil Karn) (04/10/84)
For some time now, uucico has often gotten stuck on allegra when calling out to other sites. Typically, it will wait indefinitely on a modem open call, or on a read call once the connection has been established. In doing stack traces of the code once the stuck process has been killed and dumped, the code preceeding the offending open or read appears to set alarm timeouts properly. Furthermore, the code doesn't ALWAYS fail; I've watched it many times handle timeouts from non-answering systems just as it should. It's just that alarm signals occasionally get lost. I am therefore wondering if there are some fleeting bugs in the way alarms, signals and/or setjmp/longjmp are handled in 4.2. Any ideas? Phil