[gmfawcett]

A wild nitpicker appears. I understand where you're coming from & why this matters. But Go, the language spec, doesn't make any guarantees about struct layout at all. A layout difference may be meaningful, practically, but it's potentially unreliable.

e.g. see https://groups.google.com/g/golang-nuts/c/1BlZDNBLiAM

Having said that: if a Go compiler for a given architecture decided to change its layout algorithm, I'm pretty sure it would earn a changelog entry.

[munk-a]

PHP long stated that associative array sorting order was unstable and not guaranteed (especially when the union (+) operator or array_merge function were involved) - that doesn't mean ten bazillion websites wouldn't instantly break if they ever actually changed the ordering to be unpredictable.

Language designers need to contend with the fact that the ultimate final say in whether a thing is or not is whether that behavior is observed.

[giraffe_lady]

Didn't ruby actually do exactly this though? And it broke a million websites and they changed it back in the next version and have made it explicit ever since? To me that is much stronger evidence than what we think would happen if php did it.

[vanderZwan]

I don't know about Ruby, but one example I can think of where a language made the instability explicit is that early on in the language Go changed the behavior of the select statement:

> If one or more of the communications can proceed, a single one that can proceed is chosen via a uniform pseudo-random selection.

https://go.dev/ref/spec#Select_statements

In an early implementation it would pick in lexical order, IIRC (and the specification did not mention how a communication should be picked). Not only could this lead to bugs, apparently some people were relying on it and they didn't want that.