Re: run pgindent in CI

On 11/5/26 22:34, Peter Eisentraut wrote:
> On 05.11.26 21:26, Peter Geoghegan wrote:
>> On Thu, Nov 5, 2026 at 3:01 PM Peter Eisentraut <peter(at)eisentraut(dot)org>
>> wrote:
>>> The above have been done, so here we are back on this.
>>
>> It takes me about 7 seconds to run pgindent locally, on my fast
>> workstation. I prefer to run it exactly as koel does, for obvious
>> reasons.
>>
>> Are you sure that this is an appropriate use of CI cycles?
>
> Based on my testing, 7 seconds is trivial compared to everything else
> that is going on in a CI run.  The CompilerWarnings job is regularly
> faster than all the other jobs.  We have about 5 minutes we can fill
> with stuff there before it would affect the overall experience, not that
> I'm proposing to do that.
>

+1

pgindent is one of things I regularly forget to run before pushing,
after some last minute changes and waiting for CI. And then I remember
one second after pushing. 7s is worth getting rid of that stress.

>> Nathan had
>> a patch that made pgindent run much faster. Perhaps that should be
>> committed first?
>
> I think that would be lost in the noise on CI.  It would be welcome for
> local use, of course.

I don't know how Nathan's patch works, but couldn't we keep track of
"high watermark" of commits that are pgindent-free, and then run
pgindent only on files modified by commits since then? The watermark
would need to be updated somewhat regularly, ofc.

regards

--
Tomas Vondra