Re: Maintaining a list of pgindent commits for "git blame" to ignore

Peter Geoghegan <pg(at)bowt(dot)ie> writes:
> On Thu, Mar 18, 2021 at 3:39 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> I don't object to maintaining such a file; if it makes "git blame"
>> work better, that's a huge win. However, the file as you have it
>> seems rather unreadable. I'd much rather have something that includes
>> the commit date and/or first line of commit message. Is there any
>> flexibility in the format, or does git blame insist it be just like this?

> I ended up doing it that way because I was in a hurry to see how much
> it helped. I can fix it up.
> We could require (but not automatically enforce) that the first line
> of the commit message be included above each hash, as a comment. You
> could also require reverse chronological ordering of commits. That
> would make everything easy to follow.

Given that the file will be added to manually, I think just having an
existing format to follow will be easy enough. I'd suggest something
like

b5d69b7c22ee4c44b8bb99cfa0466ffaf3b5fab9 # Sun Jun 7 16:57:08 2020 -0400
# pgindent run prior to branching v13.

which is easy to make from "git log" or "git show" output. (Because
of the precedent of those tools, I'd rather write the commit hash
before the rest of the entry.)

The date is important IMO because otherwise it's quite unclear whether
to add a new entry at the top or the bottom.

Other points: the file should be kept in src/tools/pgindent/, and
it should definitely NOT have a name beginning with ".".

regards, tom lane