overflow in C re-revisited

[utzoo!henry]

Mhtsa!rick has missed my point.  His quote from Kernighan+Ritchie,
page 185, "The handling of overflow... is machine-dependent." is
correct, but he should have read the next sentence:  "All existing
implementations of C ignore integer overflows...".  The way 99% of
all programmers interpret that second sentence is "nothing special
happens when an overflow occurs".  Wrong, as you can easily discover
if you play with it.

What the sentence really should say is "All existing implementations
of C ignore THE POSSIBILITY OF integer overflows...".  This makes it
clear that anything can happen because the compiler simply doesn't
care.  DMR and I have corresponded about this in the past and newer
manuals will presumably have revised wording.

As an aside, it would be nice if the compiler could spot overflows
in compile-time arithmetic and complain about them.  The arguments
against run-time checking don't apply there.  I realize this is
probably hard to retrofit into existing compilers, though.