bug#46834: 28.0.50; byte-compiling the standard counter closure fails

[Stefan Monnier]

>> and notice that the counter is not shared between the two functions :-(
> I'd noticed that, but figured Stefan wouldn't accept "partial byte
> compilation" as a reasonable bug scenario :-)

I'd have to ask him, but he'd probably roll his eyes and complain about
this use case yet again.

> That said, the comment in byte-compile--reify-function is incorrect:

[ I don't see which comment you're referring to.  ]

> since closures use alists and "let" uses proper lists, we can't share
> structure between them, so the return value will be equivalent to a
> snapshot of FUN, not "equal" to FUN itself. OTOH, even changing that
> wouldn't help as byte-compiled closures use a third format to store
> the bindings, IIUC.

I think we could make the shared state work if we turned

    (closure ((VAR . VAL) ...) ...)

into something like

    (cl-symbol-macrolet ((VAR (cdr ',CONSCELL)) ...) ...)

I find the idea pretty repulsive, tho.

> I've been meaning to ask, is there anything like an XFAIL test in our
> current framework? This would be an excellent use for one of those.

I don't know what is XFAIL, and I'm not very knowledgeable about our
test infrastructure, so you might want to ask elsewhere.


        Stefan