On Sun, Nov 10, 2024 at 8:12 PM Andrew Dunstan <andrew@dunslane.net> wrote:
> Oh, ouch. I guess it's too late to add those for this release. Still, an initdb failure is not as bad as a pg_rewind
failure,IMNSHO.
Yeah, I don't think it's all that bad. It seems both unlikely
(certain kinds of junction points with no intermediate directories, or
chains of junction points), and easy to work around once you get the
idea by experimentation (probably: add subdir, or tell initdb the real
target path directly, or move into place).
In hindsight I should have posted that observation onto the -hackers
thread where people able to test on Windows would have been more
likely to act, not buried here in -committers (and I didn't want to
commit blind to pre-CI branches in a hurry myself). Sorry about that.