Hi Shihao, Thanks for the review. Both points are addressed in v10.
> 1. ICU rules are not copied. Thanks for catching that. The patch is updated so the upgrade now copies template0.daticurules. I used a parameterized update so that multiline and long ICU rules are preserved as well. > 2. If a real run fails in the checks, the new cluster stays, A real --initdb run now uses the existing cleanup when compatibility checks fail, just as in --check mode. The same command can then be retried. Failures after the upgrade begins still retain the cluster. Best regards, Bohyun On Thu, Oct 22, 2026 at 5:46 AM shihao zhong <[email protected]> wrote: > > Hi Bohyun, > > Thanks for v9. My points on v6 are all handled. Two new ones. > > 1. ICU rules are not copied. If the old template0 has daticurules, > the new one has none and the upgrade still succeeds. A small fix is > attached. > > 2. If a real run fails in the checks, the new cluster stays, and the > same command then fails with "new cluster data directory is not > empty". Nothing has been copied yet at that point, so could it be > removed like in --check mode? > > Thanks, > Shihao >
v10-pg_upgrade-initdb.patch
Description: Binary data
