Re: [iptables-nft RFC 5/5] generic.xlate: make one replay test case work

[Phil Sutter <phil@nwl.cc>]

On Mon, Nov 21, 2022 at 12:19:32PM +0100, Florian Westphal wrote:
> This is just to demonstrate yet another problem.
> 
> For the rule itself it doesn't matter if '-i' or '-s' is passed first,
> but the test script has no deeper understanding for the rules and will
> do a simple textual comparision, this will fail because as-is the output
> is different than the input (options are written out in different
> order).
> 
> We either need to sanoitize the input or update the test script to
> split lines and re-order the options or similar.

My solution was to add replay records to test files like so:

| diff --git a/extensions/generic.txlate b/extensions/generic.txlate
| index 6779d6f86dec8..8c3b7dbeb7320 100644
| --- a/extensions/generic.txlate
| +++ b/extensions/generic.txlate
| @@ -1,69 +1,70 @@
| -iptables-translate -I OUTPUT -p udp -d 8.8.8.8 -j ACCEPT
| +iptables-translate -I OUTPUT -p udp -d 8.8.8.8 -j ACCEPT;-d 8.8.8.8/32 -p udp -j ACCEPT
|  nft insert rule ip filter OUTPUT ip protocol udp ip daddr 8.8.8.8 counter accept
|  
|  iptables-translate -F -t nat
|  nft flush table ip nat

Since iptables is able to compare rules though, we could utilize this.
So when checking the replay, instead of calling iptables-save and
searching the output, we could call 'iptables -C'. I'll try this, it
sounds simple and doable.

Thanks, Phil