(In reply to Iain Buclaw from comment #3) > For resolving any kind of symbol to string via CTFE, I'd suggest we defer > running semantic analysis until the second pass. > > See for instance: > https://github.com/D-Programming-Language/dmd/pull/5302/files#diff- > 0d13b271f20a2f0982c3c74062c314fdR440 Ah, it was much more trivial than that. ;-) - insn->ctfeInterpret(); + insn = insn->ctfeInterpret(); Looks like the CTFE result is not stashed in all cases.