Re: Endless pcap_stats + patch

[Guy Harris <guy () alum mit edu>]

Max Laier wrote:

> I'm not sure that's the best sollution.  For one, libpcap could choose to 
> implement pcap_stats for files later on.

Most of the statistics would be meaningless unless the capture file has 
packet drop, etc. statistics in it - libpcap format doesn't support 
that, although pcap-NG does.

However, with pcap-NG, instead of pcap_stats() returning statistics, the 
application reading the file would see a statistics record, and display 
statistics at that point.  In addition...

> Also I'd find it to be 
> surprising if I step up to a running tcpdump and ^T doesn't do anything - 
> I'd assume that something went wrong.

        1) You still get the kernel message from ^T.

	2) tcpdump isn't the only program or script that could run without 
producing any output while it's processing data - what if you're 
running, for example, a grep reading one file and writing matching lines 
to another file; ^T won't tell you anything, as grep doesn't catch SIGINFO.

	3) If it *is* useful to have tcpdump report a progress indicator, the 
indicator would just be the count of packets processed - you don't need 
pcap_stats() for that.

> A short error message to stderr 
> seems to be the right sollution, at least in my mind.

Not to mine - an error message sounds to me like an indication that either

        1) you did something wrong by typing ^T

or

        2) the program has a bug that should be fixed so that ^T works.
-
This is the tcpdump-workers list.
Visit https://cod.sandelman.ca/ to unsubscribe.