Some time ago I submitted an issue with Intel's ifort Beta version16 Fortran compiler that still does not seem to be resolved. I find that when I link a fortran code with a library that was written in C, I cannot open I/O units for formatted output. I get the error: forrtl: severe (257): formatted I/O to unit open for unformatted transfers. This error will appear no matter what unit number I try to use. I found at the time if I revert to ifort 15.0.2, the error goes away. Now, I am still getting the error with ifort version 16.0.2. Is there any hope to fix or resolve this bug?
链接已复制
Do you have a Premier Support issue number or a link to another forum thread where you reported this? I haven't been able to locate your earlier report and am not familiar with this issue. (I have written many mixed-language programs that don't have this problem.)
Also, note that your name shows here as "(name withheld)" because you have not selected a "display name" that is not an email address. You can change this by clicking on your name in the upper right and going to the Dashboard, then edit your profile.
Hi Steve, Thanks for the response. I believe I've fixed the "(name withheld issue)". Here's a link from one of the last e-mails (8/10/2015) I had from you on this issue. I presume the last number is the Premier issue number. By the way another problem I'm having is since you changed around your web-site, I haven't found a way to get back to this issue to remind myself what was done on it. If I click on this link it doesn't get me anywhere helpful. Is there still a way to get back to one's past Premier Support issue postings?
https://businessportal.intel.com/irj/portal/MyIssue2?issueNum=6000117858
Don,
Yes, you fixed the name.
You should be able to log in to Premier Support at https://premier.intel.com/ The customer side of this looks very different than the Intel side, so I am unsure what the appearance is to you or what you need to do to see your issues. I do note that the name associated with that issue isn't yours (which is why I couldn't find it), but it is clear that you were the one communicating.
I looked up that issue and the last I see is a note from you last October saying that everything was working. Are you now saying the problem has returned?
Steve,
I can log into Premier Support at https://premier.intel.com/, but don't find anyway to get from there to records of my earlier issue postings. Anyway, the problem has returned and I don't recall how it was resolved earlier, other than reverting back to ifort 15.0.2. If you can see what was advised earlier or have other suggestions that would help.
Parallel to the forensic work on your Premier Support submission, would you please post a summary of the issue here? I tried to reproduce the problem with the code below using the description in #1, and did not observe an error with Parallel Studio 15.0.5 or 16.0.2 (OpenSuse Linux, and on Windows, with "csub_" replaced by "CSUB" in the C code). My test code simply reads the source code of the Fortran part (file ptest.f90) and prints it out with line numbers added, based on a line count that it gets from the C subroutine, which obtains the count by scanning the same file.
! test I/O to text files when C subprograms in lib are linked ! https://software.intel.com/en-us/forums/intel-fortran-compiler-for-linux-and-mac-os-x/topic/628780 ! program ptest implicit none integer i,nln character(len=132) line open(11,file='ptest.lst',status='replace',action='write') call csub(nln) ! C library routine returns number of lines in ptest.f90 open(10,file='ptest.f90',status='old',action='read') ! do i=1,nln read(10,'(A)')line write(11,'(I4,2x,A)')i,trim(line) end do close(10) close(11) end program
The C subprogram:
#include <stdio.h>
void csub_(int *nln){
char lin[132];
FILE *fil=fopen("ptest.f90","r");
*nln=0;
while(fgets(lin,132,fil))(*nln)++;
fclose(fil);
}
The account you used to submit this wasn't under your name - I'm home right now and don't have access to that, but my recollection was that the email address ended in .nl. If I recall, it wasn't reproducible with 16.0 and had to do with some plotting library (PGPLOT?)
I'll look it up again when I get in tomorrow and send you an email about it.
Steve, mecej4,
As inferred from Steve's recovery of my earlier Intel Premier submission record, the I/O issue was resolved by my reinstalling/rebuilding the Plplot graphics library (the C code I was linking to). This now works fine with ifort ver. 16.
Thanks for your help, Don
