'var$2004' is being used without being defined

[SCHWOERTZIG_T_]

Hi all,

I am encountering a strange message related to run time check.
I compiled with ifort 12.1.3.293 build 20120212 on linux64 with following options : -g -O0 -fp-model precise -openmp  -Dmkl -ftz -extend-source -DFCCI -check uninit -fpe0 -ftrapuv  -traceback -lefence

Running my program leads to :

forrtl: severe (193): Run-Time Check Failure. The variable 'mysource_$ICUT' is being used without being defined
Image              PC                Routine            Line        Source             
e_mainline_linux6  0000000003B9E426  mysource_                  371  mysource.F

Line 371 is a ENDDO statement. ICUT is clearly defined with this DO/ENDDO block :
344 : DO ICUT=1,NBCUT
...
371 :  ENDDO ! next ICUT

Seting ICUT=0 at the begining of the subroutine leads to :

forrtl: severe (193): Run-Time Check Failure. The variable 'var$2004' is being used without being defined
Image              PC                Routine            Line        Source             
e_mainline_linux6  0000000003B9E426  mysource_                  371  mysource.F

But String "2004" does not occur in any source file.

Thanks for any help.
Thierry

[Kevin_D_Intel]

We have received other reports of this type of error and compiler confusion in the past that were fixed in more current releases. Your case appears to be a defect. The 12.1 is rather old release.

If you can post a complete reproducer I can verify the error and whether this has already been fixed in a newer release.

[SCHWOERTZIG_T_]

Kevin Davis (Intel) wrote:

If you can post a complete reproducer I can verify the error and whether this has already been fixed in a newer release.

In fact it seems related to a segmentation fault. Using modules leads to "-uninit" issue instead of SIGSEGV signal.
I was able to reproduce with the attached source file.

Compilation command is :
ifort source.F -g -O0 -fp-model precise -openmp  -Dmkl -ftz -extend-source -DFCCI -check uninit -fpe0 -ftrapuv  -traceback -lefence

Execution leads to :

forrtl: severe (193): Run-Time Check Failure. The variable 'idst3$DIST_PCUT' is being used without being defined
Image              PC                Routine            Line        Source             
a.out              000000000040319A  defect_IP_idst3          31  source.F

Thierry

[Lorri_M_Intel]

In this particular cutdown program, I changed "check uninit" to "check bounds" and got a bounds violation.

Oh, and yes, it gave the seemingly-bogus uninitialized message with the latest 14.0 compiler too.

Anyway, if you look in the cutdown program, the do-loop variable I runs from 1->101, but the size of that dimension is only 100.

Would you please try the experiment with your original program, to change "check uninit" to "check bounds" ?

           thank you --

                              --Lorri

[jimdempseyatthecove]

FWIW a variable that lies beyond the bounds of its dimension is also undefined. As to if this is trapped by the diagnostic code... probably a 50/50 proposition. As it would likely be using junk beyond the end of a flag table or looking for a special NaN.

Jim Dempsey