Re: [PATCH nft] tests: validate generated netlink instructions

[Florian Westphal <fw@strlen.de>]

Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> On Mon, Jul 20, 2015 at 05:10:46PM +0200, Florian Westphal wrote:
> > Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > > On Fri, Jul 17, 2015 at 03:31:39AM +0200, Florian Westphal wrote:
> > > > Don't apply -- this patch is intentionally incomplete;
> > > > I don't want to spam this list with 350k patch.
> > > > 
> > > > If you think this is worthwile to have in nft I'll push the full
> > > > changeset (this patch + 64 test files with recorded netlink debug output).
> > > 
> > > I like the idea of having more regression tests.
> > > 
> > > We can probably kill the old .t files and move the 'regression' to
> > > 'tests'.
> > 
> > Do you mean:
> > rm tests/*; mv tests/regression/* tests/ ?
> > 
> > Otherwise I'm inclined to just push what I have and let you wrangle
> > things further if needed :-)
> 
> BTW, I just remembered another bug we've got in the output.
> 
> I think we have to convert to BE the content that is printed from the
> registers, otherwise we'll have different outputs in LE and BE.

Right.  Another solution would be to keep two sets of each payload in
tests/, one for LE and one for BE platforms, and use the correct one
arcording to whatever arch the script runs on.