Description of problem: The updated 'rdate' in Fedora Core 1 adds an alarm() call to cause a 10 second timeout to occur. However, the alarm is not handled, which causes calling programs to abort up the process tree under some circumstances. Adding a trivial signal handler fixes the problem This situation occurs when the client system attempts to rdate from a server whose default input iptable policy is 'drop' (ie, blackhole). Version-Release number of selected component (if applicable): 1.3.3 How reproducible: always Steps to Reproduce: 1. block incoming rdate queries with iptables on the server side iptables -I INPUT -m tcp -p tcp --dport 37 -j DROP 2. try to rdate on the client side, querying the server 3. Actual results: The rdate call returns with a 'Alarm' message and error 142. The message isn't on stdout or stderr, which frequently blows up calling shell scripts that run from the system startup. Expected results: rdate should only return on stdout/stderr so it can be wrappered in shell scripts etc. Additional info: In rdate.c add a trivial signal handler to handle the alarm. #include <signal.h> void ALARMhandler(int sig) { exit(sig); } [...then down below their alarm(10) call...] signal(SIGALRM, ALARMhandler);
Fixed in rawhide, packages should appear there soon and will make it either in fc2 or the next release. Read ya, Phil