Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com> writes:
> We seem to be maintaining old code in pg_upgrade to not break it, see
> [3]. I can think of a fix [4] or just [5], it may not be great to do
> these kinds of fixes for all the pg_resetwal output changes. Since the
> changes we make to the output formats of controlfile or pg_resetwal
> are very rare, any of [4] or [5] looks good to me and it is better
> than backporting [2] IMO.
Yeah, the lowest-tech solution is to accept either spelling of
the description string. But I'm not very happy with the whole
approach right now. We should have an implementation that isn't
dependent on how user-friendly strings are spelled. (git seems
a bit ahead of us here, with its distinction between "plumbing"
and "porcelain" operations.)
regards, tom lane