Re: pgindent (was Re: [COMMITTERS] pgsql: Preventive maintenance in advance of pgindent run.)

Attached is a diff from git HEAD to what I get with a slightly-patched
copy of Piotr Stefaniak's latest version of FreeBSD indent. It looks
pretty darn good to me: the changes around function typedefs and
expressions containing things like sizeof() calls are alone worth the
price of admission. There are some places where it inserts extra
space because it doesn't know that a given symbol is a typedef name,
but as I remarked upthread, that's the fault of our typedef-list
collection technology not indent itself. There are not enough of those
places to make me feel like we have to have a solution right now.

There are also places where it changes its mind on how much to indent
a comment-to-the-right-of-code. I have a hack in my indent copy that
reduces the number of such diffs, but doesn't eliminate them entirely.
Removing the hack will cause indent to align such comments to the next
4-space tab stop rather than the next 8-space stop as has been its
behavior in the past. I think we should accept that change, because it
will allow squeezing a bit more text into such comments than before, but
in the interests of keeping the attached diff readable I've tried to
suppress most of those changes for the moment.

Anyway, it is now time to fish or cut bait. I don't think we can wait
much longer to decide whether we're going to adopt this new indent
version for PG 10. I think we should. The floor is open for votes.

regards, tom lane