On 2023-08-12 Sa 20:53, Peter Geoghegan wrote:
On Sat, Aug 12, 2023 at 5:20 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
Hm. I was envisioning that we should expect committers to deal
with this, not original patch submitters. So that's an argument
against including it in the CI tests. But I'm in favor of anything
we can do to make it more painless for committers to fix up patch
indentation.
I agree with this.
Making it a special responsibility for committers comes with the same
set of problems that we see with catversion bumps. People are much
more likely to forget to do something that must happen last.
After I'd been caught by this once or twice I implemented a git hook test for that too - in fact it was the first hook I did. It's not perfect but it's saved me a couple of times:
check_catalog_version () {
# only do this on master
test "$branch" = "master" || return 0
case "$files" in
*src/include/catalog/catversion.h*)
return 0;
;;
*src/include/catalog/*)
;;
*)
return 0;
;;
esac
# changes include catalog but not catversion.h, so warn about it
{
echo 'Commit on master alters catalog but catversion not bumped'
echo 'It can be forced with git commit --no-verify'
} >&2
exit 1
}
cheers
andrew
--
Andrew Dunstan
EDB: https://www.enterprisedb.com
