flang-new 19.0: "warning: overflow on power with INTEGER exponent“

[zaikunzhang]

Hi,

I am using flang-new 19.0. With the following Fortran code

! test_overflow.f90
program test_overflow
implicit none
print *, range(0.0)
print *, 10.0**(-32.0)
print *, 10.0**(-32)
end program test_overflow                                                                                                    

I got the following:

$ uname -a && flang-new --version && flang-new test_overflow.f90 && ./a.out 
Linux 6.5.0-26-generic #26~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Mar 12 10:22:43 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
flang-new version 19.0.0git (git@github.com:llvm/llvm-project.git 37c175af955f0aeab67e8c553a0a47b2ed0fdba2)
Target: x86_64-unknown-linux-gnu
Thread model: posix
./test_overflow.f90:6:10: warning: overflow on power with INTEGER exponent
  print *, 10.0**(-32)
           ^^^^^^^^^^^
 37
 1.E-32
 9.9999995E-33

Note that print *, 10.0**(-32) leads to an "overflow on power with INTEGER exponent".

What is overflowing? Is this expected or a false positive?

Thanks.

[llvmbot]

@llvm/issue-subscribers-flang-frontend

Author: Zaikun ZHANG (zaikunzhang)

Hi,

I am using flang-new 19.0. With the following Fortran code

! test_overflow.f90
program test_overflow
implicit none
print *, range(0.0)
print *, 10.0**(-32.0)
print *, 10.0**(-32)
end program test_overflow                                                                                                    

I got the following:

$ uname -a && flang-new --version && flang-new test_overflow.f90 && ./a.out 
Linux 6.5.0-26-generic #<!-- -->26~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Mar 12 10:22:43 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
flang-new version 19.0.0git (git@<!-- -->github.com:llvm/llvm-project.git 37c175af955f0aeab67e8c553a0a47b2ed0fdba2)
Target: x86_64-unknown-linux-gnu
Thread model: posix
./test_overflow.f90:6:10: warning: overflow on power with INTEGER exponent
  print *, 10.0**(-32)
           ^^^^^^^^^^^
 37
 1.E-32
 9.9999995E-33

Note that print *, 10.0**(-32) leads to an "overflow on power with INTEGER exponent".

What is overflowing? Is this expected or a false positive?

Thanks.

[klausler]

It's a bogus warning and you can ignore it. Will fix.

[klausler]

Fixed by #88188.

[zaikunzhang]

@klausler Great! Thank you for the very prompt fix.

Indeed, I spotted this bug by trying

10.0_WP**(-range(0.0_WP))

with WP being single. I hope your fix removes the warning in this case and applies to other WP as well.