On Mon, 2015-07-20 at 10:58 +0100, James Bottomley wrote:
> 
> I don't agree that all improvements, however minor are worthwhile.
> There's a cost to reviewing and merging ... that should be outweighed by
> the value of the contribution.  The scarcest thing we have is review
> bandwidth and we shouldn't waste it on minor improvements that are very
> minor. 

There is also a cost to *change*. How many times have we seen a trivial
patch which actually *breaks* something non-trivial?

We're normally quite good at managing change, and not being
conservative purely for conservatism's sake.

But trivial patches are actually the most risky in a number of ways.
They often come from inexperienced contributors, who might not spot
subtle problems introduced by naïve changes. And they are often made
without an in-depth knowledge or study of the code. The contributor
just spots a pattern (perhaps with checkpatch), and mechanically fixes
every instance they see, without stopping to look hard at each
instance.

Hell, I did this when fixing up the krealloc() users relatively
recently, getting it wrong in one case. And I really *ought* to know
better.

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse@intel.com                              Intel Corporation