[netperf-talk] code review requested

[Rick Jones]

Folks -

While running some "beat on the machine heavily" netperf omni tests 
where netperf is the reciever I am seeing a number of "error starting 
alarm timer" messages.  Some initial additions to print-out the return 
value from alarm() suggest that there was still an alarm pending (the 
return value tends to be between 1 and the value of PAD_TIME (4)).

That suggests that send_omni() in src/nettest_omni.c has one or more 
paths through it in the receive-only case where stop_timer() is not 
called like it should be.

However, I have yet to find it and am wondering if folks could take a look.

thanks, and happy benchmarking,

rick jones