On Tue, Jan 26, 2021 at 8:53 PM Michael Paquier <michael@paquier.xyz> wrote:
On Tue, Jan 26, 2021 at 10:38:43AM +0100, Daniel Gustafsson wrote: > Agreed, and pgcrypto already allows for using sha1. > > It seems like any legitimate need for sha1 could be better served by an > extension rather than supplying it in-core.
Both of you telling the same thing is enough for me to discard this new stuff. I'd like to refactor the code anyway as that's a nice cleanup, and this would have the advantage to make people look at cryptohashfuncs.c if introducing a new type. After sleeping about it, I think that I would just make MD5 and SHA1 issue an elog(ERROR) if the internal routine is taken in those cases, like in the attached.
The refactor patch looks good. It builds and passes make check.