On Wed Dec 31, 2025 at 4:26 PM CET, Tom Lane wrote:
> Ashutosh Bapat <ashutosh.bapat.oss@gmail.com> writes:
>> On Wed, Dec 31, 2025 at 5:06 PM Jelte Fennema-Nio <postgres@jeltef.nl> wrote:
>>> This tries to make running formatting a lot easier for committers, but
>>> primarily for new contributors.
>
>> I generally like the idea. Since perltidy is not enforced regularly
>> (like pgindent), running it usually ends up modifying files which are
>> not part of the patch. So I avoid it if not necessary. Do you propose
>> to make it optional?
>
> The other obstacle is that not everybody will have the right version
> of perltidy installed, but using some other version will create a
> whole lot of extraneous noise.
>
> On the whole I'd recommend not trying to automate the perltidy
> step yet. Cost/benefit is just not very good.
I would like to get to a point where it is enforced for every commit
pushed by committers, so the same as with pgindent. I agree that the
cost/benefit is not there currently, but that's what this patchset is
trying to address. The docs in the last patch try to explain clearly how
to get and configure the correct version of perltidy. And once you've
done that it's as easy as running a single make/ninja command to format
all the files.
Do you think that's not still not easy enough for committers? What would
be needed instead? Automatically fetching and installing perltidy to the
local build path when running make/ninja format?
> On the substance of the patch: I wonder whether we could make things
> more reliable by using git metadata to figure out which .h and .c
> files to point pgindent at.
I think that would definitely be helpful. The fact that pgindent runs on
files in the meson build directory when passing "." would be solved by that.
