topic Re: EOF handling in IVF vs. DVF in Intel® Fortran Compiler

EOF handling in IVF vs. DVF

salbruce — Mon, 12 Oct 2009 15:13:24 GMT

I recently converted from DVF to IVF and have had few problems, but here is one that has come up.

I'm reading a series of 8-byte strings from a text file, one string per line in the file. The list ends with an EOF mark (character 26). For example, consider the input file:

XXXXXXXX
[EOF]

In DVF the following code gives i=1, but in recompiling in IVF I get i=2 and ALIST(2) gets the EOF mark. Is there a compiler switch that will make IVF behave like DVF in this regard?

CHARACTER*8 ALIST(N) ! N is passed from calling routine
I=0
1 DO WHILE (I.LE.N-1)
READ (14,'(A8)',END=2,ERR=3) ALIST(I+1)
I=I+1
END DO

Thanks in advance.

Re: EOF handling in IVF vs. DVF

Steven_L_Intel1 — Mon, 12 Oct 2009 16:19:02 GMT

I can think of at least two different things that may be at work here, but without a full test case I'm not sure what it might be. You also don't say which "DVF" you're using.

First, early versions of DVF did not properly handle data files where the last record didn't end in a newline. That was fixed at some point fairly early in the product. Second, and I don't remember when this change was made, an end-of-file condition is not supposed to trigger an ERR= branch, though since you also have END= that may not matter.

Can you show a small but complete program that illustrates the problem?

There is no compiler switch controlling this behavior.

Re: EOF handling in IVF vs. DVF

salbruce — Mon, 12 Oct 2009 17:01:45 GMT

Quoting - Steve Lionel (Intel)

Here is a small complete example (attached, and reproduced below), along with an example input file.
I was previously using DVF6.
Thanks.

I can think of at least two different things that may be at work here, but without a full test case I'm not sure what it might be. You also don't say which "DVF" you're using.

First, early versions of DVF did not properly handle data files where the last record didn't end in a newline. That was fixed at some point fairly early in the product. Second, and I don't remember when this change was made, an end-of-file condition is not supposed to trigger an ERR= branch, though since you also have END= that may not matter.

Can you show a small but complete program that illustrates the problem?

There is no compiler switch controlling this behavior.

Re: EOF handling in IVF vs. DVF

Steven_L_Intel1 — Mon, 12 Oct 2009 17:40:11 GMT

Ok, I understand now. Yes, that's an ENDFILE record in there and I can see a difference from CVF. You can get the behavior you want by adding the /vms switch, though this should not, I think, be required. I'll let the developers know about this.

Re: EOF handling in IVF vs. DVF

Steven_L_Intel1 — Wed, 21 Jul 2010 13:56:52 GMT

This will be fixed in the next major release, due out late this year.