Florian Fainelli wrote:

I don't see why you should be creating an immutable branch for Lee and
not simply merge Rafal's "[PATCH V4 RESEND 2/2] dt-bindings: mfd: add
Broadcom's Timer-Watchdog block" patch with Lee's ack directly. This is
a new file, so I don't see how it would create conflicts as long as we
don't pile up changes on top.

Rafał Miłecki wrote:

would that be OK for you to simply ack 2/2? So Guenter can pick my
patch without the whole immutable branch & PR thing?                   

Guenter Roeck wrote:

I don't entirely see the point of that complexity for dt changes,    
but whatever. Since my tree is not the official watchdog-next tree,  
that means I can not take the entire series (which goes way beyond   
the dt changes and also drops the bcm63xx driver). Unless I hear     
otherwise, I'll drop the series from my tree for the time being      
and wait for the dt changes to be sorted out.                        

If Rob wants `dt_binding_check` to run cleanly in -next, we have to
treat the DT documentation in the same manner we do for real code
when build dependencies exist between patches.  Simply sucking them up
through a single repo is just dandy until subsequent changes are
required, which unfortunately is often the case.

Being the Maintainer of MFD, which is often the centre point of
cross-subsystems patch sets, I've been bitten by this too many times.
Hence my hesitancy to 'just Ack it and be done'.

I've been pushing back on the requirement for clean `dt_binding_check`
runs in -next for a while and would much prefer to treat it the same
way we do `checkpatch.pl`, whereby a clean run is not a hard
requirement.  Instead it is used as one of many tools to check for
inconsistencies prior to submission (as possibly against patch-sets
once they are posted onto the list).  However, just as we see false
positives in `checkpatch.pl` we should see them in `dt_binding_check`
where patches have simply been applied into different trees and may
lag each other by a week or two.

It sounded to me like Lee wanted an immutable branch for that

Not exactly, I said:

  "> Suppose we should take patch #2 via [Watchdog] as well.

   If that happens, I would like a PR to an immutable branch."

The alternative is that I take the patch and provide an immutable
branch to you, which I am in a position to do.

Of course all of this hassle just goes away if the clean
`dt_binding_check` run on -next requirement is laxed and we can just
take our own patches without fear of wrath.

-- 
Lee Jones [李琼斯]
Senior Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog