Thread

Re: pgindent versus struct members and typedefs

Nathan Bossart <nathandbossart@gmail.com> — 2026-05-06T14:44:26Z

On Tue, May 05, 2026 at 11:43:39PM -0400, Tom Lane wrote:
> Nathan Bossart <nathandbossart@gmail.com> writes:
>> For fun, I spent some time with an AI tool to develop the attached fix for
>> this problem.  The explanation seems reasonable to me, although I am by no
>> means a pgindent expert.  When I looked at this in December, I did find
>> this similar commit from upstream [0], but I failed to make the connection
>> with last_u_d.  0002 is the result of a pgindent run after applying 0001.
>> You'll notice that it fixes the exact set of cases I found with grep
>> upthread.
> 
> Those changes are clearly improvements.  I'm too tired to investigate
> right now, but I wonder if we should adopt the upstream fix you
> mention?  (Or more generally, other changes they made since we forked?)

The upstream fix is from before we forked, it just didn't fix this
particular case.  I don't see any missing changes from
pstef/freebsd_indent, but there have been a number of changes in the
FreeBSD version:

    https://cgit.freebsd.org/src/log/usr.bin/indent

Some of our changes to pg_bsd_indent bumped INDENT_VERSION.  Should we do
that here?

-- 
nathan