PATCH RFC: -fstrict-overflow

[Andrew Haley]

Eric Botcazou writes:
 > > Basically, -fwrapv extends the language standard to say "signed
 > > integers wrap."  -fno-strict-overflow does not change the language
 > > standard--signed overflow remains undefined--but the compiler does not
 > > take advantage of that fact.
 > >
 > > [...]
 > >
 > > I would be interested in hearing other opinions as to whether
 > > -fno-strict-overflow should be distinct from -fwrapv.
 > 
 > I'm of this opinion too, the semantics of the language should not
 > depend on the level of optimization, only the extent to which it is
 > exploited should.

This makes good sense.  With no-strict-aliasing we aren't defining a
different programming language; we're abstaining from performing
certain optimizations that result in incorrect programs doing
undesired things.  It's the same with -fno-strict-overflow.

Andrew.