On Mar 30, Ben Hutchings <b...@decadent.org.uk> wrote: > > > For the IPv6 and LFS legacy release goals I think it would be best if we > > > would welcome massive (automatic?) tests to find all of the outstanding > > > issues and get them fixed finally! > > I fear that the major outstanding issue is ifupdown, which basically > > does not support non-trivial dual stacked configurations and needs a > > serious redesign. > > e.g. vlan/bridging if-*.d scripts are run for every AFI. > I agree that ifupdown is broken, but how is this connected to the > IPv6 goal? For IPv6 it is at least using iproute2 and not bad old > ifconfig... The problem is with all features implemented by external if-*.d scripts. If e.g. a bridge is created by the first defined afi, the second script will fail. And if it does not fail on up then everything will still break on a down event, when the script run by the first AFI will destroy the interface and ifupdown will be able to remove the IP address of the second AFI. Bugs were opened long ago, but there is no interest/manpower to fix them (which is not surprising if you have ever looked at the ifupdown source).
-- ciao, Marco
signature.asc
Description: Digital signature