Re: [PATCH nft v3 1/2] jump: Introduce chain_expr in jump and goto statements

[Phil Sutter <phil@nwl.cc>]

Hi,

On Fri, May 24, 2019 at 11:21:48AM +0200, Pablo Neira Ayuso wrote:
> On Fri, May 24, 2019 at 09:29:34AM +0200, Fernando Fernandez Mancera wrote:
> > On 5/24/19 9:17 AM, Fernando Fernandez Mancera wrote:
> > > Hi Pablo,
> > > 
> > > On 5/21/19 9:38 PM, Fernando Fernandez Mancera wrote:
> > >> Hi Pablo,
> > >>
> > >> On 5/21/19 11:28 AM, Pablo Neira Ayuso wrote:
> > >>> On Thu, May 16, 2019 at 10:45:58PM +0200, Fernando Fernandez Mancera wrote:
> > >>>> Now we can introduce expressions as a chain in jump and goto statements. This
> > >>>> is going to be used to support variables as a chain in the following patches.
> > >>>
> > >>> Something is wrong with json:
> > >>>
> > >>> json.c: In function ‘verdict_expr_json’:
> > >>> json.c:683:11: warning: assignment from incompatible pointer type
> > >>> [-Wincompatible-pointer-types]
> > >>>      chain = expr->chain;
> > >>>            ^
> > >>> parser_json.c: In function ‘json_parse_verdict_expr’:
> > >>> parser_json.c:1086:8: warning: passing argument 3 of
> > >>> ‘verdict_expr_alloc’ from incompatible pointer type
> > >>> [-Wincompatible-pointer-types]
> > >>>         chain ? xstrdup(chain) : NULL);
> > >>>         ^~~~~
> > >>>
> > >>> Most likely --enable-json missing there.
> > >>>
> > >>
> > >> Sorry, I am going to fix that.
> > >> [...]
> > > 
> > > I am compiling nftables with:
> > > 
> > > $ ./configure --enable-json
> > > $ make
> > > 
> > > And I am not getting any error, am I missing something? Thanks! :-)
> > > 
> > 
> > Fixed, the option is --with-json. Why isn't it "--enable-json" as other
> > features?

It is actually not that uniform. While we have:

--enable-debug
--enable-man-doc
--enable-python

we also have:

--with-mini-gmp
--with-cli
--with-xtables
--with-json

and all of them just enable/disable something, unlike --with-python-bin.

> We can just update this to accept both, either --with-json or
> --enable-json.

For consistency, we should turn all of the above --with flags into
--enable ones, but there's of course the compatibility problem.

What do you think, is it feasible to change all the above and introduce
--with aliases for them?

Cheers, Phil