[rbldnsd] ipv6 support: beginnings of an ip6trie dataset

[Jeff Dairiki]

On Sun, Apr 28, 2013 at 07:49:39PM +0400, Michael Tokarev wrote:
> The new thing is just excellent.  I really really like it, especially
> the way how it performs much better than my ip4trie code, -- much
> faster and uses less memory.  Fun.

Thank you, again.
 
> Please take a look - how do you like my changes...

A couple of comments, both minor:

1. Regarding patch "2c6ba85 asserts are actually ENabled by default"

No, I don't think that's right.  Asserts are currently disabled by
default.  If you run configure in a clean directory, NDEBUG will be
defined in config.h.

This is (IMHO) as it should be.  The default configuration should be
one appropriate for deployment.  Just as, by default, CFLAGS contains
-O, the default should be asserts disabled.

(I've just reverted this in my ipv6 branch.

   https://github.com/dairiki/rbldnsd/commit/d58a52b
)


2. Regarding "66f69ab add python tests to DISTFILES and move it to main dir"

(And this is a very minor nit.)  To avoid cluttering the source, I
personally liked keeping the python test code in its own subdirectory.
But if you want to go with the flat directory structure, I would
suggest renaming 'rbldnsd.py' to something else (to avoid it getting
lost/confused with all the other rbldnsd* files).

The main program is called rbldnsd.  There is an rbldnsd.c and an
rbldnsd.py.  Were I just starting to suss out the code, I would
definitely be confused by this.

Perhaps 'testlib.py'?

(If you'd like me to do the renaming, I'd be happy to.  Let me know.)

> and let's release the damn thing :)

Fire at will!

> Please note: I haven't tested actual IPv6 code, especially on uncommon
> architectures.  I probably should do that before actual release ;)

More testing can't hurt, but I'm not too worried about it.  The only
code that is most likely to exhibit architecture-dependent bugs is the
stuff in btrie.c.  The ip4trie dataset exercises that code just as
well as the ip6trie dataset.

Cheers,
Jeff