On 13.07.21 20:26, Tom Lane wrote:
> Did you see my followup? The vast majority of live systems do not do
> that, so we are accomplishing nothing of value by insisting it's a
> crash-worthy bug.
But there are no guarantees that that will be maintained in the future.
In the past, it has often come back to bite us when we relied on
implementation-dependent behavior in the C library or the compiler,
because no optimization might invalidate old assumptions.
In this particular case, I would for example be quite curious how those
alternative minimal C libraries such as musl-libc handle this.
