Hello, Tomasz I decided that CALM is very important thing and do not disturb thou from it. In final state it will be definitely fasm 1.x killer.

As I understand: now is no needance to split encodings(on main & alter) - flags like rmdest|rmsrc accesible from outer and they taken into account in encoding.

Tomasz, can thou add in x86 one more element espFixer which metadata will be affected by "push", "pop", "sub esp,imm", "add esp, imm" natively in encoding. And make by default if something related to esp it is actualy related to esp + espFixer, and modernize all instruction: if one of its members related to espFixer - in expression make summ imm part with espFixer metadata value, while expressions that consist from only esp and imm stay unchanged.
espFixer is my name. Thou could name flag for stack top change amount tracking as thou want.
Thanks if it is realizable.

It will be not problem to preserve or manualy change that metadata in places that are branched, or where assumed that stack changed indirect (via ret from callee that restores stack) and so on places.

I welcome any interesting ideas like these. It may still be a few months before I start working on "fasm 2" CALM package, but in the meantime I can collects ideas and think about architecture. The "x86-2.inc" was just an early prototype.

But I would also like to stress out that with fasmg's way of implementing instruction sets as easily tweakable headers, there is always an option of making additional packages for specific purposes. Now with CALM's processing power projects like my assembly of legacy sources could become a legitimate add-ons allowing to assemble syntax of some ancient assemblers, for example.