Re: [PATCH iproute2-next] tc: pedit: Support JSON dumping

[Petr Machata <petrm@mellanox.com>]

Petr Machata <petrm@mellanox.com> writes:

> Stephen Hemminger <stephen@networkplumber.org> writes:
>
>> On Sun, 26 Apr 2020 12:23:04 -0600
>> David Ahern <dsahern@gmail.com> wrote:
>>
>>> On 4/23/20 3:59 AM, Petr Machata wrote:
>>> >
>>> > Stephen Hemminger <stephen@networkplumber.org> writes:
>>> >
>>> >> On Wed, 22 Apr 2020 20:06:15 +0300
>>> >> Petr Machata <petrm@mellanox.com> wrote:
>>> >>
>>> >>> +			print_string(PRINT_FP, NULL, ": %s",
>>> >>> +				     cmd ? "add" : "val");
>>> >>> +			print_string(PRINT_JSON, "cmd", NULL,
>>> >>> +				     cmd ? "add" : "set");
>>> >>
>>> >> Having different outputs for JSON and file here. Is that necessary?
>>> >> JSON output is new, and could just mirror existing usage.
>>> >
>>> > This code outputs this bit:
>>> >
>>> >             {
>>> >               "htype": "udp",
>>> >               "offset": 0,
>>> >               "cmd": "set",   <----
>>> >               "val": "3039",
>>> >               "mask": "ffff0000"
>>> >             },
>>> >
>>> > There are currently two commands, set and add. The words used to
>>> > configure these actions are set and add as well. The way these commands
>>> > are dumped should be the same, too. The only reason why "set" is
>>> > reported as "val" in file is that set used to be the implied action.
>>> >
>>> > JSON doesn't have to be backward compatible, so it should present the
>>> > expected words.
>>> >
>>>
>>> Stephen: do you agree?
>>
>> Sure that is fine, maybe a comment would help?
>
> Something like this?
>
>                         /* In FP, report the "set" command as "val" to keep
>                          * backward compatibility.
>                          */
> 			print_string(PRINT_FP, NULL, ": %s",
> 				     cmd ? "add" : "val");
> 			print_string(PRINT_JSON, "cmd", NULL,
> 				     cmd ? "add" : "set");

I just sent it as a v2 of the patch, we can discuss there.