svn commit: r328257 - in head/sys: arm/broadcom/bcm2835 dts/arm modules

[Ian Lepore]

On Sat, 2018-01-27 at 13:32 -0800, Rodney W. Grimes wrote:
> > 
> > On Sat, 27 Jan 2018 12:13:57 -0800
> > Adrian Chadd <adrian.chadd at gmail.com> wrote:
> > 
> > > 
> > > Hi,
> > > 
> > > Find the middle ground. Don't dissuade the developer too much.
> >  This is what happened two years ago when I started hacking on the
> > allwinner SoCs :
> > 
> >  - I asked what should be done for bringing a new board
> >  - andrew@ told me that we first need to switch to upstream dts and
> > update drivers.
> >  - Guess what, I did that.
> Great, thats good co-operatation and communications, sometimes though
> it is not so smooth.  The better we become at dealing with the not
> so smooth the faster forward progress can be made.
> 
> > 
> > > 
> > > Here's an example:
> > > 
> > > Make the driver follow DTS, allow a tunable/kenv check for it to
> > > override whether it needs to be in the DTS or not (the "keep phk happy
> > > for now" compromise) and have it default to obeying the device tree.
> > > 
> > > That way phk is kept happy and the defaults are the same as normal-ish
> > > ARM /and/ you have a springboard to experiment with extending FDT
> > > overlays at runtime for people who wish to do so.
> >  I don't care about keeping phk@ (or any other developer) happy, we
> > have a standard, let's stick to it.
> *sigh*  Let me ask you if you do not care about keeping any other
> developers happy, why should any of them be concerned about keeping
> you happy?  We need to always try to find middle ground and
> co-operate in positive ways.
> 
> On the "we have a standard" front, well when standards get in the way
> of forward progress they are often side stepped.  Maybe this standard
> is not such a good standard and warts are going to form around it. I
> have seen some discusssion at least on ways to improve the current
> situation, hopefully someone takes them and runs with them.
> 
> Others have pointed out they do not like the current model in that
> it gets in the way of developement progress.   I can see this point.
> I can see phk's points, and I can see your points.
> 
> IMHO if we shove the standard down our own throats we are in
> effect cutting our hands off in the process, not somethig we
> really want to do is it?
> 
> > 
> > > 
> > > (I personally hate having to edit the dts/recompile/reboot for every
> > > test hardware change; it makes breadboarding things up kinda
> > > hilariously annoying.)
> >  Use overlays then. And if you don't want to reboot provide patch for
> > loading overlays at runtime.
> Are those the only solutions?

Lots of nice platitudes, but the bottom line is that the commit
violates the existing conventions in ways that make it the only such
violator.  The reason isn't "because it's the only way to make
progress", the reason is "because I'm philosophically opposed to
needing a reboot after making a one-time configuration change to how
the hardware will be used."

Hacks to move forward in the face of broken or missing functionality
are something we have to live with, given how understaffed we are for
correcting the deficiencies.

Hacks to assuage someone's personal preferences or philosophy, IMO,
should not be something we have to live with.

-- Ian