Treatment of machine integers (esp. in Python version)

[AndrewTolmach]

p. 3 para -1 says that an $int$ in $\mathcal{L}_{int}$ is a 63-bit signed integer.

• In fact, AFAICT all the assembly code generated prior to Ch. 10 assumes 64-bit signed integers.

• Python actually does not parse a leading '-' as part of an integer (it parses as a unary negation even if there is no separating space), so in the concrete syntax of $\mathcal{L}_{int}$ , $int$ should really be a positive integer only. (I guess after, e.g., partial evaluation, we need the abstract syntax to be able to store a signed int.) This is a small point, but may confuse students.

• More seriously, none of the support code (quoted in the book or otherwise) actually performs machine arithmetic (modulo $2^{64}$) on the source language or in the X86 interpreter. This really undercuts the nice diagram (1.8)! At the very least, this deviation deserves a footnote of explanation. I also have patches that provide a small set of utility routines for doing machine arithmetic in Python and invoking them from the interpreters, checkers, and the input_int routine (which should clamp to machine range in order to match the behavior of scanf in runtime.c).

[jsiek]

The patch for doing machine arithmetic in Python would be much appreciated! Could you do a pull request with that?