volatile vs. noalias

[Doug Gwyn]

In article <14797 at mimsy.UUCP> chris at mimsy.UUCP (Chris Torek) writes:
>... it is not true of `noalias', at least as it was conceived and proposed.
>It could only be ignored when not used, not even by library routines called
>by routines called by you.

Yes, we didn't get it specified right.  But even some of the most ardent
opponents of "noalias" privately admitted that the specification could
probably have been fixed.  It just wasn't feasible to do that at that
stage of the standardization effort.  It's something the next C standards
committee should keep in mind (although perhaps global intramodular
optimization will be common by then, obviating the need for it).