[Swan-dev] libevent loose ends

[D. Hugh Redelmeier]

I've finished a review of the libevent branch.  Here are some more
comment.

It looks to me as if the code implements --retransmit-timeout as a bound 
on a single retransmission delay, not the sum of all retransmissions 
delays for this message.  This does NOT match the documentation.

struct event's get allocated like crazy.  Does it make any sense to reuse 
them?  The ones associated with interfaces look to be re-used.

Do we have leaks?

The current compiler warnings might indicate a leak or at least a
loose end.  I haven't looked carefully.


I see reapchildren is #if 0.  Are we finished with subprocesses?  If so, 
delete the thing.


The new flags are --retransmit-timeout and --retransmit-interval.
The corresponding things within the code are not named consistently
CD_RETRANSMIT_T, CD_RETRANSMIT_I,
.r_timeout, .r_interval
These seem particularly well-designed to confuse the reader.
r, T, and I are not at all suggestive.