Re: An independent IPv6 patching of bsd-finger
On Thu, 11 Feb 2010, Mats Erik Andersson wrote:
Now I can report on Fedoras patching. First off, their patch is from 2005,
but the youngest code change is from 2001, originating in the USAGI project.
The Fedora location is
http://cvs.fedoraproject.org/viewvc/rpms/finger/devel/bsd-finger-0.17-usagi-ipv6.patch
A simplified difference file, containing all relevant C-code, is appended
to this message.
It turns out that the USAGI patch uses practically the same technique as
I myself used in the second form of my interdiff, but there are small
differences that need to be brought into light and to be discussed.
1. USAGI uses code insertion that needs activation by a macro INET6
in order to be included. My suggestion does not introduce a conditional
code alteration. At this point in time IPv6 should be mandatory, from
my view point.
Is there someone in Fedora we can bring into this conversation? Maybe the
maintainer there?
To minimize Debian's delta, perhaps we can accept the conditional code
alteration for now, and just make sure to set that flag as we build the
package.
Perhaps you we can run with their USAGI-based diff, but add our own small
changes. Then we can ask upstream to accept our delta, which is the
USAGI-based diff in Fedora plus your changes.
2. For protocol determination, USAGI uses PF_UNSPEC, I use AF_UNSPEC.
It seems from my quick reading that AF_UNSPEC is better, although they
happen to share the same constant. I'm no expert at this bit, though!
[rest snipped]
My conclusions are threefold:
A. The compiler test on the macro INET6 should be avoided.
B. Debian should use AI_ADDRCONFIG.
C. Debian could either keep gethostbyname() in fingerd/fingerd.c,
or can follow USAGI in using getaddrinfo() also there, but should
then also add AI_ADDRCONFIG, thus going beyond USAGI.
I think these are right, mostly. With regard to the compiler test, I
think it's okay for now, since accepting it minimizes our delta to
someone else's code.
Thanks for checking out what other people are doing!
Have you checked in with upstream and started a discussion with the
relevant Fedora maintainer? If not, please do so and keep this list (or at
least that bug) in the loop!
-- Asheesh.
--
Any resemblance to real persons, living or dead, is purely coincidental.
Reply to: