@f52985, I just noticed that this older PR is still awaiting your attention as well. :)

Oops, I didn’t even realize this PR existed. 😅

The issue is that the new functions $sof and $store are translated into bodies containing only a single instruction: Return.

This happens because of an assumption in the translator based on the specification:

Whenever a function returns store, it is assumed to be used in a store-passing style, for example:

-- if s1 = f(s)
-- if s2 = g(s1)
-- if s3 = h(s2)
...

In this pattern, each function call is isolated, and the return value of each call is bound to a single variable.

Based on this assumption, the IL2AL translator removes explicit store passing and uses the variable s as if it were the single variable used to access the global store.

However, the new functions $sof and $store do not follow this assumption. Instead, they are used in an inline style, where the function call appears as a subexpression within another expression, and removing the return statement brings the wrong result.

Because of this, I believe these functions require special handling at some point—either through a hint in the specification, during translation, or during AL interpretation.

My choice was to add a hardcoded case in the interpreter that treats $store specially, forcing it to always return the global store.