Editorial: more explicit validation in CreateDynamicFunction

[bakkot]

Fixes #2373. cc @erights

[erights]

I am not a reviewer, but FWIW, LGTM. Thanks much!

[syg]

I understand where @erights is coming from in #2373, but I find this piecemeal parsing, then re-parsing the entire text only for the Early Error application to be misleading. It's misleading in that it suggests that parsing expr using exprSym alone is somehow insufficient to do all the validation while it in fact is.

I'd prefer to keep to a single parse and have a NOTE that says the syntactic constraint that @erights wants highlighted is applied post-concatenation. Then again, I don't find that state of affairs particularly surprising or indirect as he does.

[bakkot]

I think this might actually be normatively necessary - that is, I think #2348 might accidentally not have been editorial after all.

Consider new Function('/*', '*/ ) {'). This shouldn't be legal, I believe that #2348 accidentally made it so.

[erights]

@bakkot Nice example! I tried to come up with one but did not. Thanks!

[syg]

Consider new Function('/', '/ ) {'). This shouldn't be legal, I believe that #2348 accidentally made it so.

Ah ha, I retract my above comment then.

[syg]

lgtm with comment

[erights]

I like it. But your example is worth a thousand such explanations since most of us didn't see how the original text was erroneous even after we started looking for a counter-example. Your example is tiny and clarifying. It would be a good addition.

[bakkot]

Added my above example as another NOTE.