On Wed, May 29, 2019 at 01:24:19PM +0000, Hamlin, Garick L wrote:
> I ran clang checker and noticed these. It looks like the
> sha2 implementation is trying to zero out state on exit, but
> clang checker finds at least 'a' is a dead store.
>
> Should we fix this?
> Is something like the attached sensible?
> Is there a common/better approach to zero-out in PG ?
This code comes from the SHA-2 implementation of OpenBSD, so it is not
adapted to directly touch it. What's the current state of this code
in upstream? Should we perhaps try to sync with the upstream
implementation instead?
After a quick search I am not seeing that this area has actually
changed:
http://fxr.watson.org/fxr/source/crypto/sha2.c?v=OPENBSD
--
Michael