Allow CTE on basic type/math functions

[mvorisek]

as requested in #10771 (comment)

once this PR is merged, we can discuss the remaining functions

[mvorisek]

@TimWolla @nielsdos can you please review?

[iluuu1994]

Thanks!

[nielsdos]

I went through the PR again. I think the functions on which you added CTE are safe.
I only have 2 additional remarks:

1. For some functions, like number_format() and variadic ones, the benefit might be only very small. That's because the optimizer limits the amount of arguments processed in CTE calls to 3. This is done for simplicity, and for compile-time performance (a lot of works happens for CTE stuff, so it makes sense to limit this).
2. I see there's some discussion about the CTE & refcount attribute ordering. IIRC someone told you to reorder this and now it is requested that the reordering is undone. I don't know what the best way forward is here. I'd just look at the most commonly used ordering and use that in new code, while keeping the old code unchanged to avoid extra code churn.

[mvorisek]

For some functions, like number_format() and variadic ones, the benefit might be only very small. That's because the optimizer limits the amount of arguments processed in CTE calls to 3. This is done for simplicity, and for compile-time performance (a lot of works happens for CTE stuff, so it makes sense to limit this).

As long as a function is CTEable, I would say there should be no arbitrary limit of arguments.

I see there's some discussion about the CTE & refcount attribute ordering.

Yes, someone wants it, someone not. I would say 6 total changes across the whole php-src are worth vs. the total "preferred" order.

[mvorisek]

Let's please merge this PR if there is no feedback left so I can finish finish another PR on top of it. Thank you.

[nielsdos]

It seems that first compile-time-eval followed by refcount is the most common form. So therefore I think making this consistent in this PR is fine.
The functions marked as CTE were determined to be correctly marked.

[iluuu1994]

Can is_resource cause issues when called at compile time? Otherwise, it could still eliminate calls when elements are not resources.

[mvorisek]

Correct me if I am wrong, but can resource be even observed on CT?

[iluuu1994]

No, but it could eliminate calls like is_resource(42).

[mvorisek]

I added is_resource, tests pass, the 1 ASAN failure is unrelated.