[PATCH] Document arithmetic overflow semantics

Richard Kenner kenner@vlsi1.ultra.nyu.edu
Thu Feb 13 15:45:00 GMT 2003


    I believe this generalization is drifting away from the initial
    specific topic.

Sure, but for what I consider a good reason: if we're going to treat
*certain* undefined behavior differently than others we need to:

(1) Be explicit that we are doing that; and
(2) Precisely characterize which behaviors are in each set.



More information about the Gcc mailing list