Re: [PATCH blktests v3 3/3] nvme: introduce nvmet_target_{setup/cleanup} common code

[Daniel Wagner]

On Fri, Aug 25, 2023 at 07:34:21AM +0000, Shinichiro Kawasaki wrote:
> IMO, SC2119 is not broken. SC2119 (and its companion SC2120) assumes that bash
> functions do not have optional arguments. If any functions which refer arguments
> are called without arguments, it complains. With the assumption, SC2119 is
> useful to detect missing arguments of function calls. (I guess Bart thinks this
> is useful.)

I wanted to say that the implementation of SC2119 is broken, not the
SC2119/SC2120 itself. Sorry for the confusion.

> However, when we implement argument parsers in bash functions so that the
> arguments can be optional, the assumption for the SC2119 is wrong. Then SC2119
> reports are useless. Until recently, blktests had few functions with such
> optional arguments, such as _init_null_blk or _init_scsi_debug. But most of
> calls to those functions had some arguments, and it was rare to call them
> without any argument. So SC2119 reports were easy to suppress and not a pain.
>
>     I doubt Shellcheck can be improved and detect if functions have the optional
>     argument parsers...

No idea. But we are not alone with this problem:

https://github.com/koalaman/shellcheck/issues/2511

> Recently, you actively cleans up tests/nvme/* (which is great!), and introduced
> argument parsers in test/nvme/rc. The first one is _nvme_connect_subsys, and the
> second one is this _nvme_target_setup. It looks for me this is a bash coding
> style change in blktests, from "don't use optional arguments often" to "use
> optional arguments aggressively".

Yes, it's a bit excessive to hand in all possible arguments all the
time. Especially it makes it even hard to review if only value changes
but 6 default values are passed to the setup function.

> If we apply this change, we should suppress
> SC2119. If we keep the old coding style, we should keep on enabling SC2119. What
> I see here is the style difference between you and Bart.
>
> Now I'm tempted to disable SC2119, and to go with the new coding style...
> 
> If I have any misunderstanding, or if anyone has more comments on this, please
> let me know.

All good from my side.