See discussion in end of #216 .
4-to-6 lookup in stateful NAT64 with empty pool4 will fail for point-to-point interfaces because the point-to-point address is used instead of the local interface address.
I see. You are completely right; in fact, the code from empty.c's __pick_addr() (the one that chooses the address in the 6->4 direction) outputs ifa_local instead of ifa_address. This confusion was a major oversight on my part; I apologize.
I can confirm that replacing all mentions of ifa_address into ifa_local does not break anything in the environment where I originally tested the code. I'd encourage you to propose a pull request so Git will credit you for your work, but if you are not interested I'll commit the change at some point late next week (and release 3.4.4).
Great, posted a PR. Are you able to estimate the schedule for the next release? Wondering if I should switch to my own repo for our deployment or wait for the new release.
(or did you mean it is scheduled for late next week?)
Since the PR is already available and there are no other changes since 3.4.3 I can start the release preparations right away. If the full test suite doesn't yield any problems (I'm optimistic, as this was a very specific tweak), I should be able to release this happily tomorrow.
Note, however, that #216 itself is part of Jool 3.5.0, which is expected to come out in about a month (unless the code review yields another trainwreck bug. We are currently facing one of those).
We're owing you three fixes since you started #216 (216, the incorrect FAQ statement and this). Do you want additional credits in Jool's README? If so, please respond (or add to the PR) the information you'd like included.
Great! I can work around #216, no rush.
If you want to add my name (Andreas Urke) and a link to my github account I would be honoured 👍 I'm happy I am able to contribute something back