tl;dr I am happy to redo the patches.
On 6/30/21 1:12 AM, David Newall wrote:
> I rely on diff to see what is different between two versions, and not
> necessarily
> consecutive versions. It is an immensely effective technique for
> finding where
> something became broken.
I also rely on that. I agree completely it is immensely effective.
I split the change in two commits precisely because I also care about
legible diffs. But I recognize there are many opinions on how best to
achieve that. I am happy to try it another way.
> Diffs like that which John proposes add noise, dramatically increasing
> the effort
> required to see real changes. Did he simply change indent or did he
> also change
> text? I can't tell at a glance.
The nature of the change is putting a decent chunk of existing autoconf
m4 within a newly-introduced if-then-else. That means either the
indention will become wrong, or there will be some churn/noise
reindenting the old code.
I chose to reindent first, making on odd 2x indent, and then add the new
code and if-then-else, also fixing the indent.
Another option would be to switch the order of the two patches: making
the meaningful change without touching the old code, and then indenting
the old code another level. A third option would be to simply do the
first of those to patches, making the minimal change and then leaving
the indentation incorrect.
I am happy to resubmit the patches broken down according to either of
those 2 alternatives, or any other alternative.
John