[llvm-dev] How to disable UBsan divide-by-zero in source files?

Eli Friedman via llvm-dev llvm-dev at lists.llvm.org
Mon Mar 30 10:01:19 PDT 2020

https://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html#issue-suppression .

Also, the float-divide-by-zero diagnostic is no longer part of -fsanitize=undefined, as of clang 9.0.


-----Original Message-----
From: llvm-dev <llvm-dev-bounces at lists.llvm.org> On Behalf Of Jeffrey Walton via llvm-dev
Sent: Sunday, March 29, 2020 4:18 PM
To: llvm-dev <llvm-dev at lists.llvm.org>
Subject: [EXT] [llvm-dev] How to disable UBsan divide-by-zero in source files?

Hi Everyone,

I'm catching a finding for divide by zero when using floats. I think
this is a false positive since IEEE 754 states either a trap or

    $ cat test.c
    #include <float.h>
    #include <math.h>
    int main(void)
      return INFINITY == 1.0f / 0.0f ? 0 : 1;

    $ clang -fsanitize=undefined test.c -o test.exe
    $ ./test.exe
    test.c:5:27: runtime error: division by zero

I'm looking for a way to disable UBsan divide-by-zero in the source
code. I don't see something like:

     #pragma clang diagnostic ignored "-fsanitize=divide-by-zero"

Trying to use a pragma results in

    $ clang -fsanitize=undefined test.c -o test.exe
    test.c:4:34: warning: pragma diagnostic expected option name (e.g.
    #pragma clang diagnostic ignored "-fsanitize=divide-by-zero"

How do disable divide-by-zero at the source level?

The use case is, users who run self tests will
CFLAGS="-fsanitize=undefined". They should not need to investigate
findings, and add additional flags like no-divide-by-zero. Things
should just work for them.

Thanks in advance.
LLVM Developers mailing list
llvm-dev at lists.llvm.org

More information about the llvm-dev mailing list