FWIW:
My experiences with -fstrict-aliasing is that it doesn't generate a warning
every time there is an actual problem. This may be a source of hair loss if
you are trying to correct all aliasing issues. I can't point to a concrete
example or test case, but I do remember in my own projects having to
manually find the issue because GCC didn't. Also, there is a lot of uses
with the union of a float and an int -- these are technically illegal (see
3rd paragraph of http://davmac.wordpress.com/2010/02/26/c99-revisited/). I
resolved all of my aliasing between float <-> int by using memcpy() or a
simple asm function:
float I2F(int x); ->
flds $4(%esp)
ret
On Wed, Jun 30, 2010 at 5:51 AM, Michael Menegakis <arx...@gmail.com> wrote:
> Thanks. I also went and made the compiler warn about it.
>
> Places where "warning: dereferencing type-punned pointer will break
> strict-aliasing rules [-Wstrict-aliasing]" occurs on gcc 4.6.0:
>
> code/qcommon/net_ip.c
> code/renderer/tr_backend.c
> code/renderer/tr_shade.c
> code/renderer/tr_shade_calc.c:
> code/renderer/tr_surface.c:
>
> On Wed, Jun 30, 2010 at 1:09 PM, Patrick Baggett
> <baggett.patr...@gmail.com> wrote:
> > Any time a float and int are used with aliased points, for short.
> > e.g.
> > int* x = (int*)&float_value;
> > *x = 0x7F000000; //set to 1.0f or something
> >
> > union asdf {
> > int x;
> > float f;
> > } var;
> > var.f = 123.0f;
> > var.x >>= 2; //aliasing issue.
> >
> > See C99 spec for what is/isn't legal. Check network and file code for
> > problems.
> _______________________________________________
> ioquake3 mailing list
> ioquake3@lists.ioquake.org
> http://lists.ioquake.org/listinfo.cgi/ioquake3-ioquake.org
> By sending this message I agree to love ioquake3 and libsdl.
>
_______________________________________________
ioquake3 mailing list
ioquake3@lists.ioquake.org
http://lists.ioquake.org/listinfo.cgi/ioquake3-ioquake.org
By sending this message I agree to love ioquake3 and libsdl.
