[Dnsmasq-discuss] [bisected] re-reading resolv.conf bug

[Simon Kelley]

On 03/01/2022 19:30, Johannes Stezenbach wrote:
> Hi Simon,
> 
> commit efea282396 "Fix logic in add_update_server() to make
> optimisation actually optimise." introduced a bug when re-reading
> resolv.conf after changes. Found by git bisect and confirmed
> by reverting efea282396 on top of current git 011f8cf1d011.
> 
> The issue happens when starting/stopping VPN and entries are
> added to/removed from /run/dnsmasq/resolv.conf.
> --strict-order is in use.
> 
> 
> No VPN:
> nameserver 192.168.178.1
> 
> After starting VPN:
> nameserver 10.x.xx.60
> nameserver 10.x.xx.75
> nameserver 192.168.178.1
> 
> Jan  3 20:06:04 dnsmasq[20581]: using nameserver 192.168.178.1#53
> Jan  3 20:06:04 dnsmasq[20581]: using nameserver 10.x.xx.75#53
> Jan  3 20:06:04 dnsmasq[20581]: using nameserver 192.168.178.1#53
> 
> If dnsmasq is restarted to work around the problem,
> then after stopping VPN:
> 
> Jan  3 19:55:18 dnsmasq[18927]: using nameserver 10.x.xx.60#53
> 
> 
> Could you please check it?
> 
>

I did check it and it's a bit of a mess. Reverting efea282396 only
coincidentally fixes the problem, knowing that it does made finding the
real problem much easier, so thanks for that information.

I believe the real fix is at

https://thekelleys.org.uk/gitweb/?p=dnsmasq.git;a=commit;h=553c4c99cca173e9964d0edbd0676ed96c30f62b

Which sorts things, to the best of my knowledge. Please could you check
and see if it sorts things for you too?


Cheers,

Simon.