Thanks ambrus. You summed up the situation exactly.

It was actually really hard to remember a situation where the evaluation order affected the outcome of the statement. I've been bitten by it a couple of times over the years. I've taken part in a couple of threads where it was the cause of a problem or otherwise the subject of discussion.

It means that each time I write any kind of a compound statement, or when I look over someone elses code here, looking for the source of their problem, I have to stop, look and think hard about whether there is some kind of evaluation order dependancy involved. Often, I will break compound statements into two parts, frequently needing to add a temporary variable, just so I can try it, to see if it makes any difference. Mostly it doesn't, but I'm sure it did once a long while ago.

And that's the point. Perl allows, and even encourages the use of compound statements. Like 'em or hate 'em, Perl allows them and they do have utility.

Sometimes they can clarify. Or simplify. Or reduce the chance of errors by making it harder for a single compound action to be interspersed accidently. Sometimes they produce an elegant solution (I wish I could find the discussion from around 3 years ago where Abigail-II demonstrated his triple assignment trick!?).

But as is, with the nebulous existance of the possibility of an evaluation order dependancy, using compound statements always leaves a doubt in my mind, and one which I find extremely hard to erase through knowledge or practice.

I think that given the rarity with which Perl code could actually benefit in any meaningful way from the possibility of the optimisations that an undefined evaluation order affords, I'd prefer that the loophole was quashed and my doubts eradicated.

This is one Perl 5 ambiguity that could and should be closed.