Strange Error

[SoundUnbound]

Hello. I am getting a very strange error. I am using a Visual Fortran 90 Ver 6.0 compiler and am getting a sqrt domain error, but I know that all the values being input to the square root are positive. I put a print statement before the declaration as a sanity check, and it printed a positive value and the program kept running as if there wasn't a problem. Then I removed the print statement, got the same error, then reinserted it, no error. I reinstalled the compiler, and still had the same problem. Any thoughts?

Adrian

 

[ArkM]

As usually it's not a compiler error, it's this program error (99.999...%. Present the code fragment at (and immediatly before) 'bad' sqrt call.

 

[SoundUnbound]

Sorry for taking such a long time to respond. Code fragment before and after:

print *, 'check'
dydx(1)=(d_hat/y(1))*(avg_r+y(1)**2*avg_r_4/(pi*d_hat*x))**0.5)
*(1 + 2*B_hat/(3*y(1)))**(-1)*(C_i/C_0 - avg_r_p/avg_r_4)
END SUBROUTINE

The variable of interest here is x. After further debugging, I found that a previous subroutine was passing an infinite value to x. But how can inserting a print statement prevent this from occuring? Btw, I am using an ODE solver from Numerical Recipes, if that sheds any light.

Regards

 

[xwb]

No idea why your error is happening but the expression could be optimized.

! instead of
(avg_r+y(1)**2*avg_r_4/(pi*d_hat*x))**0.5
! use
sqrt (avg_r+y(1)**2*avg_r_4/(pi*d_hat*x))

! instead of
(1 + 2*B_hat/(3*y(1)))**(-1)
! use
1.0/(1 + 2*B_hat/(3*y(1)))

The print statement may prevent a crash but is the result reliable? What happens if you

print *,'check ',x

 

[SoundUnbound]

Thanks for the tips xwb. Do you know why making those adjustments optimizes the code? I tried the print command you suggested, but it didn't seem to change anything. I am writing a lot of data to files, could that trigger this problem?