Re: [PATCH v1] treewide, serdev: change receive_buf() return type to size_t

[Johan Hovold <johan@xxxxxxxxxx>]

On Fri, Dec 15, 2023 at 06:07:33PM +0100, Francesco Dolcini wrote:

> On Fri, Dec 15, 2023 at 05:18:52PM +0100, Johan Hovold wrote:
> > On Fri, Dec 15, 2023 at 02:55:59PM +0100, Francesco Dolcini wrote:
> > > To me the change is correct, with that said probably this should have
> > > been explicitly mentioned in the commit message or a separate
> > > preparation patch.
> > 
> > It's a separate change and should not be hidden away in a tree-wide
> > change that goes through a different maintainer.
> > 
> > Please drop this change from this patch and resubmit it separately to me
> > if you want and I'll review when I have the time.
> 
> Fine, I agree.
> 
> I see those options (let me know if you see other options I have not
> mentioned):
> 
> 1. I add this change (taking into account also intel ice) as a separate
>    patch in this series and you may just ack it and Greg could merge
>    together with the serdev one.
> 2. I prepare an independent patch for the GNSS change and only once this
>    is merged I'll send a rebased v2 of this one.
> 3. I update this patch without this GNSS API change, that mean I will
>    have to cast away the signed type from a few GNSS drivers.
> 
> 1 is my preferred option, 2 is fine, but it seems a little bit of overdoing,
> 3 I would avoid, we are doing this cleanup to be a little bit more
> strongly typed and to prevent the kind of bugs that is the original trigger
> for this patch.

Changing the return type of gnss_insert_raw() is going to be a bit more
involved and should be done in a separate patch (e.g. you need to look
at gnss_usb_rx_complete() and ice_gnss_read() to avoid introducing new
warnings there).

And both option 2 and 3 will introduce conversion warnings (W=3, which
we have plenty of anyway) unless you add casts.

I suggest you go with 3, unless you insist on 2.

Johan