| Issue |
172499
|
| Summary |
[clang] __builtin_isgreater/__builtin_isgreaterequal/__builtin_isless/__builtin_islessequal/__builtin_islessgreater does not handle unordered comparisons
|
| Labels |
clang
|
| Assignees |
|
| Reporter |
zatrazz
|
The following code triggers an invalid exception on ABIs that suppors 128 bit float types (either by long double or extended types like __float128):
$ cat t.c
#include <math.h>
#include <stdio.h>
#include <fenv.h>
int main (int argc, char *argtv[])
{
feenableexcept (FE_INVALID);
long double qNaN = __builtin_nanl ("");
long double zero = 0;
int r = __builtin_isgreater (zero, qNaN);
printf ("%d\n", r);
}
$ clang -target aarch64-linux-gnu t.c -o t -D_GNU_SOURCE -lm
$ ./t
Floating point exception (core dumped)
It is not a problem for compiler-rt because it does not support floating point exceptions; but is an issue when using the gcc libgcc. This is triggered by the glibc testsuite when either using clang as the test compiler [1] or by using clang as the built compiler [2].
My plan is to disable an optimization on glibc math.h where isfunction is routed to its builtin conterpart, as least for types different than float and double.
[1] https://sourceware.org/bugzilla/show_bug.cgi?id=33322
[2] https://sourceware.org/git/?p=glibc.git;a=commit;h=be48ae12b10b0556895e356c3ae34b8ca229e61b
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
