character function with variable length

[Intel_C_Intel]

Hi everybody,

I am new to FORTRAN 90 and try to test the (good) changes since FORTRAN 77. In wanting to use character functions with length calculated at run time I ran into a problem with the compiler. I do not now whether this is due to the compiler (I am using a relatively old version: 5.0D) or whether my code is wrong. I have stripped down the code to this:

character(8) char,ch

char='bonjour'
ch=f(adjustl(char))

contains

function f(string)
character(*), intent(in) :: string
character(len(string)) f

f=string
return
end function

end

and still have an 'assertion failure' message. I read and read the code but see nothing wrong according to the FORTRAN90 manuals I have with me. Please could somebody:

- signal me any gross error,
- or tell me whether this code compiles with a more recent version of Compaq Visual Fortran.

Thanks in advance.

[Steven_L_Intel1]

It compiles with CVF 6.5.

Steve

[Jugoslav_Dujic]

Works fine for me at 5.0D also -- pasted from here?! Where do you get
assertion failure -- at compile-time or at run-time?

Jugoslav

[Jugoslav_Dujic]

Works fine for me at 5.0D also -- pasted from here?! Where do you get
assertion failure -- at compile-time or at run-time?

----------------(few minutes later)---

Oh, it works only in Debug version -- right, it crashes in Release, even with
/optimize:1

Jugoslav

[Intel_C_Intel]

Some more details,

Within the large program the code comes from, I have the same behavior. The Release version crashes at compile time while the Debug version passes through (Jugoslav, I had not tried /optimize:1). However, it is worth knowing that I have a strange behavior at run-time with the Debug version. At some point I get a Microsoft Developer Studio message:

"User breakpoint called from code at 0x77f862e8".

If I click OK the program seems to work fine. Does this message make sense to somebody? Is this real bad?

[Jugoslav_Dujic]

To clarify: there is no problem with your code, but with the compiler. "User breakpoint called from XXX" messages can be ignored, since as far as I know those are some kind of checkouts in run-time libraries; (kind of ASSERTs?).

By the way, my DVF 5.0D when F95 compatibility warnings are turned on warns that "Variable-length character functions are obsolescent in F95". Is it really so? I don't see a reason why they are declared obsolete?

[Steven_L_Intel1]

The "user breakpoint" message is not a "problem" in the compiler - this is how the Run-Time Library signals I/O errors. You should get a real error message in the console window.

Assumed-length (CHARACTER*(*) functions are indeed obsolescent in Fortran 95. Not our doing, and we'll continue to support them. See my Dr. Fortran article in Visual Fortran Newsletter III.

Steve

[Intel_C_Intel]

Thanks to you guys, I now know that the problem is with the compiler, not the code. Also that the code compiles with CVF6.5 and I am seriously considering upgrading. However, I am still wondering whether the bug has been truly fixed. Compiling doesn't mean running (see my previous message). I suppose that for people with an inside knowledge of the compiler it is not too difficult to identify the bug in the short code submitted. Hence my question:

Steve, has this bug been identified and fixed?

Roger

[Steven_L_Intel1]

I believe so, yes. We found and fixed several problems in this area over the past few years.

Steve