Re: IndexTupleDSize macro seems redundant

On 2018-01-11 13:26:27 -0500, Tom Lane wrote:
> I wonder whether there is a way to get alignment traps on Intel-type
> hardware.  It's getting less and less likely that most hackers are
> developing on anything else, so that we don't see gotchas of this
> type until code hits the buildfarm (and even then, only if the case
> is actually exercised in regression testing).

It's possible, but a lot of code out there, including things like glibc,
assumes that unaligned accesses are fine. Therefore it's very painful to
use.

Greetings,

Andres Freund