Matt Bauman

Similarly, I think a descriptive warning (like Alex's suggestion) could be back ported to 1.6/1.7, but a prescriptive one should not be.

> I'd also point out creating [a tool] is straightforward Folks have been saying this _for years_ while simultaneously saying that `using` is fine (e.g., whenever someone comes onto Discourse...

I don't see this as changing anything about how we use semver; I see this as changing **responsibility** for something we've deemed to be ok (and will continue to be...

We’re almost perfectly in agreement. I agree this is what should happen. But it’s not what has been happening. I would further clarify my prior comment on responsibility to say...

I **really** don't like this, but thank you. My previous — obviously wrong — understanding was that the Julia compiler was trying to guard us from LLVM's undefined behaviors via...

That matches my expectations, though. It's just fine that `r()` uses an old value of `a` and `f()` uses a new one and that the two may not be consistent....

I don't disagree with _any_ of the things in this PR being marked as unspecified or implementation-defined or flat-out invalid. What I dislike is the term UB and the privilege...

Clearly, I know. That const replacement crash is a good example, thanks. But it still feels different to me than how UB has traditionally been _exploited_ by other compilers. Maybe...

> `reduce` and `mapreduce` be added to the list of implementation-defined behavior No, I think that's a different sort of thing; every single function needs to specify the behaviors its...

> I don't know why people keep interpreting what I'm saying as trying to add UB everywhere That's not at _all_ what I'm doing here. This whole thing started because...