Re: [RFC PATCH net-next mlxsw 03/14] selftests: forwarding: README: Document customization

[Petr Machata <petrm@nvidia.com>]

Jakub Kicinski <kuba@kernel.org> writes:

> On Tue, 26 Mar 2024 11:31:31 +0100 Petr Machata wrote:
>> Jakub Kicinski <kuba@kernel.org> writes:
>>
>> > a standard feature of kselftest. If "env" file exists in the test
>> > directory kselftest would load its contents before running every test.
>> >
>> > That's more of a broader question to anyone reading on linux-kselftest@
>> > if there's no interest more than happy to merge as is :)
>> >
>> > On Mon, 25 Mar 2024 18:29:10 +0100 Petr Machata wrote:
>> >
>> >> +The variable NETIFS is special. Since it is an array variable, there is no
>> >> +way to pass it through the environment. Its value can instead be given as
>> >> +consecutive arguments to the selftest:
>> >> +
>> >> +	./some_test.sh swp{1..8}  
>> >
>> > Did you consider allowing them to be defined as NETIF_0, NETIF_1 etc.?
>> > We can have lib.sh convert that into an array with a ugly-but-short
>> > loop, it's a bit tempting to get rid of the exception.  
>> 
>> Maybe we could do this though?
>> 
>> 	NETIFS="swp1 swp2 swp3 swp4 swp5 swp6 swp7 swp8" ./some_test.sh
>> 
>> But NETIFS is going to be a special case one way or another. That you
>> need to specify it through several variables, or a variable with a
>> special value, means you need to explain it as a special case in the
>> documentation. At which point you have two exceptions, and an
>> interaction between them, to describe.
>
> I think there's some value in passing all inputs in the same way (thru
> env rather than argv). I guess it's subjective, you're coding it up, 
> so you can pick.

I kinda like the NETIFS="a b c" approach. If somebody wants to code that
up, I'll be happy to review :) I might get around to it at some point.