Re: [edk2-platforms][PATCH 1/8] Platform/RPi: Add model family detection

["Ard Biesheuvel" <ard.biesheuvel@linaro.org>]

On Mon, 18 Nov 2019 at 19:32, Pete Batard <pete@akeo.ie> wrote:
>
> On 2019.11.18 18:05, Leif Lindholm wrote:
> > On Mon, Nov 18, 2019 at 05:58:05PM +0000, Pete Batard wrote:
> >> On 2019.11.18 17:51, Leif Lindholm wrote:
> >>> On Thu, Nov 14, 2019 at 04:07:33PM +0000, Pete Batard wrote:
> >>>> From: Samer El-Haj-Mahmoud <samer@elhajmahmoud.com>
> >>>>
> >>>> Add GetModelFamily to RASPBERRY_PI_FIRMWARE_PROTOCOL.
> >>>>
> >>>> This uses the board revision to return a numeric value representing
> >>>> the RPi family (1=RPi, 2=RPi2, 3=RPi3 and 4=RPi4).
> >>>>
> >>>> Knowing the Pi family will help us set the SD card routing when we
> >>>> introduce support for the Pi 4 and should also be easier to maintain
> >>>> than if using individual model detection.
> >>>>
> >>>> Also add a missing entry for the "Raspberry Pi Compute Module 3+" in
> >>>> RpiFirmwareGetModelName ().
> >>>
> >>> Can you drop the above line and include the below as 1/? in v2?
> >>
> >> Okay.
> >>
> >> Note that since you requested alphabetical for PCDs, I'm going to have an
> >> "Also" in 2/ (now 3/) since the existing PCDs in
> >> Platform/RaspberryPi/Library/PlatformLib/PlatformLib.inf are out of
> >> alphabetical order.
> >
> > Actually, I try to never request reordering of existing lines, so I
> > would be quite happy for you to skip the changes that would motivate
> > the use of the "also".
> >
> > I tend to apply a rule of trying to insert *new* (or moved) lines in a
> > way that will improve the existing order - or in messy cases at least
> > not make it worse.
> >
> > I have had it pointed out to me that this is maybe not entirely
> > obvious...
>
> Well, this is exactly what I would point out as an example of the strive
> for commit atomicity getting in the way of a more readable codebase as
> well as overall user experience (the users here being the developers who
> are dealing with the code). The reason I'm pointing this out is that, in
> the past, I have been dealing with projects that seemed to care more
> about keeping a squeaky clean commit history than they seemed to care
> about making the underlying code as good as it could possibly get, which
> resulted in increased pain for the developers having to contend with
> said codebase and ultimately end-users of the software produced from
> that codebase.
>
> Again, I would assert that there has to exist a middle ground between
> keeping a super-clean commit history and improving the source where it
> can indeed be improved at little cost, by not always defaulting to
> people having to devote extra time splitting patches.
>
> But I understand this is not my choice to make here. Thus I'll stay away
> from reordering that doesn't have to do with new PCDs being introduced.
>

Please keep in mind that when open source maintainers take ownership
of your code, they assume the responsibility to ensure that it doesn't
get broken by future updates elsewhere in the codebase, often way
beyond the commercial lifetime of the product that is supported by
that code. This is a sizable effort, and an important part of managing
that effort is ensuring that the code is in an acceptable shape to
begin with, and what 'acceptable' means differs between different
maintainers. Not being able to revert a patch easily because it
touches unrelated code may make our lives more difficult years after
you have stopped caring about this platform entirely.