Re: Preliminary results for proposed new pgindent implementation

Andres Freund <andres(at)anarazel(dot)de> writes:
> On 2017-06-16 13:34:01 -0400, Tom Lane wrote:
>> I do intend to apply the diffs to HEAD in multiple steps, just to
>> make them more reviewable. But I think we should probably absorb
>> all the changes we want into v10, not leave some for later cycles.

> Btw, how much are you planning to backpatch these?

Well, that's something we need to discuss. I originally argued for
back-patching the new rules, whatever they are (ie, run the new
pgindent on the back branches whenever we've agreed that the dust
has settled). But I'm starting to realize that that's likely to
be horrid for anyone who's carrying out-of-tree patches, as I know
a lot of packagers do for instance. We have to trade off our own
inconvenience in making back-patches against inconvenience to
people who are maintaining private patchsets.

One idea that occurs to me after a few minutes' thought is to
announce that we will reindent the back branches, but not till
around the time of v10 final release. Once v10 is out, anybody
who's carrying a private patchset will be needing to think about
rebasing it on top of reindented code anyway, so dealing with that
in the back branches at the same time might be a bit less work.

Or we could leave the back branches alone and anticipate five
years worth of pain in back-patching. I don't find that very
appetizing personally, but it might be the easiest sell to the
majority of the community, since very few of us do back-patching
work on a regular basis.

regards, tom lane