NDT-DEV email list created

[Aaron Brown <>]

Hey Peter,

Yeah, if you could add the SSL_shutdown call, that'd be good. Other than that, I'm fine with it being merged.

Cheers,

Aaron

On Nov 17, 2015, at 1:59 PM, Peter Boothe <> wrote:

Closing the loop on this:

Would you like me to change the close_connection() to call SSL_shutdown() ?  Also, are there any other changes you would like before this can be merged?

  -Peter

On Fri, Nov 13, 2015 at 7:03 PM, Aaron Brown <> wrote:

Hey Peter,

On Fri, Nov 13, 2015 at 5:30 PM, Peter Boothe <> wrote:

shutdown_connection does SSL_shutdown/SSL_free, but close_connection just does
the SSL_free. Should close_connection be doing SSL_shutdown too? It'd be nice
if those could somehow be merged, but given there's only one user of shutdown,
and it's unlikely to need changed much, that's probably not a big deal.

My understanding is that SSL_free() behaves like close(), but SSL_shutdown() behaves like shutdown(). So I used one where we had the one and the other where we had the other.

The shutdown_connection does both SSL_free and SSL_shutdown, whereas close_connection just does SSL_free. From looking at the SSL_free man page, I think it's good form to do SSL_shutdown and then SSL_free no matter what, but I'm not sure.

 

Looking at this set of changes, I think it changes the snapshot semantics:

-      lastThroughputSnapshot->time = secs() - tmptime;
+      lastThroughputSnapshot->time = secs() - start_time;

I think tmptime got updated with the end of the previous snapshot, so the
'time' field will be the duration of that snapshot. With the above change, the
time will be the duration from test start till end of the snapshot.

Looking at the old code, tmptime was set once before the loop at line 387 and then is never set again until after the loop on line 448 where it is changed to be the elapsed time.  During the loop, I think tmptime remains unchanged? If I am reading things wrong, please let me know and I will update my code.

Sigh, I wondered if that was the case. Looking back at the original code, the semantics appear to be that each snapshot has the throughput covering the entire test so far, which doesn't seem useful to me as that would just mask the per-snapshot network performance. 

Cheers,

Aaron

--

ᴹ̶LAB | Measure the Internet, save the data, and make it universally accessible and useful.