On Sat, Dec 26, 2020 at 06:16:37PM +0900, Michael Paquier wrote: > The CF bot at http://cfbot.cputube.org/ includes tests on Windows, so > those problems would have been detected beforehand. Did you look at > these? If this cannot be fixed, could it be possible to revert > please? It looks rather clear that this has not been tested across > multiple platforms, and the absence of tests to allow the buildfarm to > stress this code does not really help either in gaining confidence > that this is stable.
I have been working on this patch for almost two months, and wanted to try to get it into the tree near Christmas as sort of a Christmas present to the community. It has been a tough year, and I know there are a many users waiting for this feature, even though these commits only get us a small way to the goal. I also felt the tree would be quiet so if I broke it, the disruption would be minor. What I did learn from this is that the thing I was most concerned about, the crypto code and the OpenSSL API changing between OpenSSL versions, was a non-issue, but I forgot to test for the no-OpenSSL case, forgot to do more spellchecking, and didn't check the TAP tests. Windows was more of a minor issue than I thought. The only guaranteed user-visible feature for PG 14 is the ability to have ssl_passphrase_command prompt from the terminal when started via pg_ctl. All other parts of this patch series will have to be disabled/hidden by the time we get to PG 14 beta unless we can get the data encryption part into the tree before then. -- Bruce Momjian <[email protected]> https://momjian.us EnterpriseDB https://enterprisedb.com The usefulness of a cup is in its emptiness, Bruce Lee
