On 08/26/2016 01:12 PM, Michael Neuffer wrote: > netcfg does not ask for a pointopoint address when entering the > network config manually, not even in expert mode > > In expert mode it will show you, after collecting all the information, > that is has NOT set a pointopoint address. > > Please change this setup since this always means that you have to > select "No network configuration" and manually write your > /etc/network/interface file after the installation of the base system. > > netcfg should in normal and expert mode always ask for the pointopoint > adress and let the users know that it is OK to leave the field emtpy > when there is no pointopoint address needed. > > Many large hosters use setups where you have to set a pointopoint > address and not supporting it makes a Debian installation > unnecessarily cumbersome.
Right now it only uses this scheme if it detects a certain type of network interface (by name, not even by actual type). The state machine skips netmask and gateway in this case as well. So this would require: * A way to pass in the debconf priority to netcfg_get_pointopoint (low in your case instead of critical). * A helper function to encode the list of plip/slip/ctc/escon/iucv as p2p devices. Use that from GET_IPADDRESS and use it in GET_POINTOPOINT to jump the correct next state. * GET_GATEWAY does a check if the gateway makes sense (netcfg_gateway_reachable). I suppose if that function returns unreachable, we could have an option to state that it should be reached via a host route. Maybe check for the priority or if it's actually preseeded? Otherwise still show the current error? * Probably we need to hop into GET_POINTOPOINT after GET_GATEWAY in some fashion. The previous bullet point already alludes to that. Kind regards Philipp Kern
signature.asc
Description: OpenPGP digital signature