This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PR 23572 warnings for out of range floating-point constants


On 20/03/07, Richard Henderson <rth@redhat.com> wrote:
On Wed, Mar 14, 2007 at 06:13:38PM +0000, Manuel López-Ibáñez wrote:
> +  /* We also give a warning if the value underflows.  */
> +  if (REAL_VALUES_IDENTICAL (real, dconst0)
> +      && !REAL_VALUES_IDENTICAL (realvoidmode, dconst0))
> +    warning (OPT_Woverflow, "floating constant exceeds range of %qT", type);

"exceeds" is misleading here.


ex·ceed 1. To extend beyond or outside of: The river exceeded its banks. 2. To be greater than; surpass: "a horror of the spirit that cannot be exceeded at the hour of birth or death" Robert Louis Stevenson. 3. To go beyond the limits of: exceeded my allowance. See Synonyms at excel.

I think 1 and 3 correspond to the intended meaning here but if you
have a better suggestion... I am not a native speaker.

Also, I think you should be changing the real.c routines a bit here,
as 1e-1000000000 won't be detected, since that underflows inside the
string parser.


Are you suggesting that we emit a warning from the string parser? Do we do this for any other cases? I can give it a try.

Any suggestions about my concerns regarding the testcases (testing for
infinity support)?

Thanks,

Manuel.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]