Re: HAVE_FAST_UNALIGNED_ACCESS

[Eli Zaretskii <eliz@gnu.org>]

> From: Mattias Engdegård <mattias.engdegard@gmail.com>
> Date: Sat, 1 Apr 2023 09:42:53 +0200
> Cc: Po Lu <luangruo@yahoo.com>,
>  vibhavp@gmail.com,
>  rpluim@gmail.com,
>  emacs-devel@gnu.org
> 
> 1 apr. 2023 kl. 08.39 skrev Eli Zaretskii <eliz@gnu.org>:
> 
> > I don't want to make changes on the release branch without a positive
> > evidence that the existing code causes verifiable problems.  And even
> > then I prefer to make changes only for platforms where those problems
> > happen.
> 
> I'm not overly worried about the code, but to put this matter to rest, what about simply dropping the vectorisation altogether, on all platforms, in Emacs 29? That should be very safe.

From my POV, the safest code is the one we have, because that have
lived for many moons and have been used by many people, with no
problems reported so far.  Which is why I said above that I want to
see verifiable problems with the existing code, and understand those
problems well enough, before I will agree to such non-trivial changes
on the release branch

IOW, "best engineering practices" are for master; on the release
branch, using existing code proven by time takes precedence, because
our ability to predict consequences is limited.