Fix GH-9769: Improve error message for array unpacking in constant expression

[jhdxr]

I didn't attach a test script for this as it's only changing the error message. But I can add one if it's required.
(fix #9769 )

[iluuu1994]

I didn't attach a test script for this as it's only changing the error message. But I can add one if it's required.

Yes please 🙂 They're not required per-se but are certainly welcome.

[jhdxr]

I didn't attach a test script for this as it's only changing the error message. But I can add one if it's required.

Yes please 🙂 They're not required per-se but are certainly welcome.

If this is the case I prefer to explore to bring Traversables to const expr, but this definately needs to change the target branch to 8.2 or master. Just adding a test case for error message seems to be kinds of overkilling IMHO.

[iluuu1994]

@jhdxr Since this would involve calling user code from constant expressions at least a mail to the mailing list would be warranted. This also likely has suffers from the same caching issue as this RFC did.

[jhdxr]

As said in #9769 I'll jus leave this PR open for RMs to decide.
Test case will NOT be added since error messages are subjected to change.

Exploration on Traversables in constant expression is another topic and I'll send an email to interals if I have progress on that.

[iluuu1994]

Test case will NOT be added since error messages are subjected to change.

I don't think user code should depend on error messages. It's fine for us to depend on them though, as there's no other way to test them. For bulk-changes we have the --bless parameter.

[cmb69]

@ramsey is fine with changing the error message, so this can go into PHP-8.1.

However, I think we should have a test for this; not because of the error message, but rather because this code path is apparently not tested so far (otherwise the changed message should have triggered a test failure). @jhdxr, could you please add a simple test (basically, just const EXAMPLE = [...new ArrayObject()]; for instance).

[jhdxr]

@cmb69 @ramsey test added. sorry for the long waiting.

[jhdxr]

the AppVeyor failure seems to be unrelated, but I can rerun it if necessary.

[cmb69]

Thank you! Besides the minor issue pointed out below, this looks good. And yeah, that AppVeyor failure is unrelated; I've retriggred the build nonetheless.

[cmb69]

Thank you!