Re: Rework the way multixact truncations work

On November 8, 2015 11:52:05 AM PST, Noah Misch <noah(at)leadboat(dot)com> wrote:
>On Sun, Nov 08, 2015 at 11:11:42AM -0800, Andres Freund wrote:
>> On November 8, 2015 12:54:07 AM PST, Noah Misch <noah(at)leadboat(dot)com>
>wrote:
>>
>> >I have pushed a stack of branches to
>> >https://github.com/nmisch/postgresql.git:
>> >
>> >mxt0-revert - reverts commits 4f627f8 and aa29c1c
>> >mxt1-disk-independent - see below
>> >mxt2-cosmetic - update already-wrong comments and formatting
>> >mxt3-main - replaces commit 4f627f8
>> >mxt4-rm-legacy - replaces commit aa29c1c
>> >
>> >The plan is to squash each branch into one PostgreSQL commit. In
>> >addition to
>> >examining overall "git diff mxt2-cosmetic mxt3-main", I recommend
>> >reviewing
>> >itemized changes and commit log entries in "git log -p --reverse
>> >--no-merges
>> >mxt2-cosmetic..mxt3-main". In particular, when a change involves
>> >something
>> >you discussed upthread or was otherwise not obvious, I put a
>statement
>> >of
>> >rationale in the commit log.
>>
>> I'm not following along right now - in order to make cleanups the
>plan is to revert a couple commits and then redo them prettyfied?
>
>Yes, essentially. Given the volume of updates, this seemed neater than
>framing those updates as in-tree incremental development.

I don't like that plan. I don't have a problem doing that in some development branch somewhere, but I fail to see any benefit doing that in 9.5/master. It'll just make the history more convoluted for no benefit.

I'll obviously still review the changes.

Even for review it's nor particularly convenient, because now the entirety of the large changes essentially needs to be reviewed anew, given they're not the same.

---
Please excuse brevity and formatting - I am writing this on my mobile phone.