> On 4 Oct 2021, at 21:54, Tom Lane <t...@sss.pgh.pa.us> wrote: > > Daniel Gustafsson <dan...@yesql.se> writes: >>> On 4 Oct 2021, at 15:56, Tom Lane <t...@sss.pgh.pa.us> wrote: >>> I used to think it was better to go ahead and manually reflow, if you >>> use an editor that makes that easy. That way there are fewer commits >>> touching any one line of code, which is good when trying to review >>> code history. However, now that we've got the ability to make "git >>> blame" ignore pgindent commits, maybe it's better to leave that sort >>> of mechanical cleanup to pgindent, so that the substantive patch is >>> easier to review. > >> Yeah, that's precisely why I did it. Since we can skip over pgindent sweeps >> it >> makes sense to try and minimize such changes to make code archaeology easier. >> There are of course cases when the result will be such an eyesore that we'd >> prefer to have it done sooner, but in cases like these where line just got >> one >> word shorter it seemed an easy choice. > > Actually though, there's another consideration: if you leave > not-correctly-pgindented code laying around, it causes problems > for the next hacker who modifies that file and wishes to neaten > up their own work by pgindenting it. They can either tediously > reverse out part of the delta, or commit a patch that includes > entirely-unrelated cosmetic changes, neither of which is > pleasant.
Right, this is mainly targeting comments where changing a word on the first line in an N line long comment can have the knock-on effect of changing N-1 lines just due to reflowing. This is analogous to wrapping existing code in a new block, causing a re-indentation to happen, except that for comments it can sometimes be Ok to leave (as in this particular case). At the end of the day, it's all a case-by-case basis trade-off call. -- Daniel Gustafsson https://vmware.com/