[cfe-dev] Which floating point to bool conversions should trigger warnings?
Hal Finkel via cfe-dev
cfe-dev at lists.llvm.org
Fri Apr 22 15:57:36 PDT 2016
----- Original Message -----
> From: "Richard Trieu via cfe-dev" <cfe-dev at lists.llvm.org>
> To: "cfe-dev" <cfe-dev at lists.llvm.org>
> Sent: Friday, April 22, 2016 5:47:59 PM
> Subject: [cfe-dev] Which floating point to bool conversions should trigger warnings?
> I recently added (r267054) and then revert (r267234) some warnings
> for floating point to bool conversions pending more discussion on
> the topic. From my experience, bool conversions are tricky and a
> source of many bugs, so a warning in this place would be good.
> Bool behaves a little differently than other integer types. For
> instance, 0.5 is converted to zero for integer types, except for
> bool which it gets converted to true. Also, bool is the type for
> conditionals and resulting type of logical operators.
> However, at least one style guide says to use floating point to bool
> conversion (
> https://webkit.org/code-style-guidelines/#null-false-and-zero )
I don't see where it says that.
Regardless, I think we should always warn.
> With that, I hope to get a little more discussion on this topic
> before implementing the floating point to bool conversion warnings
> again. The two factors for the warning are the source of the
> floating point value and where the conversion happens, although
> there may be other factors we need to take into account. Only the
> source has been taken account so far since it is the easier way to
> Exact floating point literals (0.0, 1.0)
> Rounded floating point literals (0.5, 4.0)
> Exact compiler time constant (5.0 - 4.0, kZero)
> Inexact compiler time constant (1.0 / 2.0, kHugeNumber)
> Run time values (getFloat())
> Function Argument
> Assigning/Initializing bool variable
> Return statement of bool returning function
> Condition of if statement, for loop, while loop, or do-while loop
> Condition of conditional operator (?:)
> Operand of logical operators(&&, ||, !)
> Currently, exact floating point literals are not warned on, rounded
> floating point literals are under -Wliteral-conversion, and
> everything else is under -Wfloat-conversion.
> Note that exclusionary groups may also be useful, for instance
> "-Wfloat-conversion -Wno-some-bool-warning" if that helps filter out
> some of the more noisy warnings.
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory
More information about the cfe-dev