[oxura]

Follow-up to #1714.

A presence guard used to bring its own FID target into the build, so it was always true. Preprocessor queries no longer root entries. #define bodies still do: a macro-expanded call must keep its target.

The new runtime fixture covers absent/present guards, continued/commented directives, C digraphs, and a macro call across a multiline comment. It failed in clang before the fix and now builds and prints 42.

The JS twins now use FID(imul) to dispatch to the real Math.imul, instead of returning a constant. The C twins still invoke the Bend entry.

Locally, all four affected/boundary fixtures pass checkup, the interpreter, emitted JS, and native C. No private-cluster run claimed.

Repo gate: 54/54; all caps unchanged.

[Lorenzobattistela]

Thanks for following up on the #1714 note. We're going to close this one: nothing in the repo uses FID on a preprocessor line, and the only effect today is that an unused def stays in the build (the guard is always true). That doesn't justify a second regex plus per-line directive tracking in c_ids while comp.ts is at its 66k cap, and since #1439 reshaped c_ids, a rebase would mean rebuilding the loop anyway. If a real effect source ever needs an FID presence guard, we can revisit it then.