Traceback contains all "Unknown" Routines, Lines and Sources

[jfdannen]

I am trying to get traceback information for my program, yet the traceback always contains "Unknown" for Routine, Line, and Source. To figure out what is going on, I created ovf.f90 from online and have compiled using "ifort -O0 -fpe0 -traceback ovf.f90 -o ovf.exe" (as directed). One other thing: I also get the message "Message Catalog System: corrupt file." three times just before the stack trace.

[Ron_Green]

You will also need -g on the compilation/link to get symbolic information in the stack trace. -traceback simply enables the traceback mechanism.

I'll look into the "message catalog...", seems I've seen that recently as well.

ron

[jfdannen]

I already tried that. Below is a copy of the Terminal window that shows exactly what I did. -JFD3

dannenho% cat ovf.f90
program ovf
real*4 x(5),y(5)
integer*4 i

x(1) = -1e32
x(2) = 1e38
x(3) = 1e38
x(4) = 1e38
x(5) = -36.0

do i=1,5
y(i) = 100.0*(x(i))
print *, 'x = ', x(i), ' x*100.0 = ',y(i)
end do
end
dannenho% ifort -O0 -fpe0 -traceback -g ovf.f90 -o ovf.exe
dannenho% ovf.exe
x = -1.0000000E+32 x*100.0 = -1.0000000E+34
Message Catalog System: corrupt file.Message Catalog System: corrupt file.Message Catalog System: corrupt file.forrtl: error (72): floating overflow
Image PC Routine Line Source
ovf.exe 00059277 Unknown Unknown Unknown
ovf.exe 00058713 Unknown Unknown Unknown
ovf.exe 0002F78C Unknown Unknown Unknown
ovf.exe 0000F226 Unknown Unknown Unknown
ovf.exe 0001606A Unknown Unknown Unknown
libSystem.B.dylib 908CE5EB Unknown Unknown Unknown
Unknown FFFFFFFF Unknown Unknown Unknown
ovf.exe 00001862 Unknown Unknown Unknown
ovf.exe 0000181C Unknown Unknown Unknown
ovf.exe 00001749 Unknown Unknown Unknown
Unknown 00000001 Unknown Unknown Unknown
Abort

[Ron_Green]

The lack of symbolic information in the -fpe0 traceback is a bug that has been reported. This bug is being worked.

As for the 'message catalog ...' message, this is something we see on Tiger but not on Leopard based systems. This involves the setting of NLSPATH environment variable in our compiler source script, ifortvars.sh. Specifically, lines 22-26 of this script:

22 if !($?NLSPATH) then
23 setenv NLSPATH /opt/intel/fc/10.1.014/lib/locale/en_US
24 else
25 setenv NLSPATH /opt/intel/fc/10.1.014/lib/locale/en_US:${NLSPATH}
26 endif

The message does not affect the operation of the compiler. However, it is annoying. To get rid of this message you can safely remove lines 22-26 in ifortvars.sh. This will be fixed in a future 10.1 compiler update.

ron

[jfdannen]

Actually I am using Leopard (and compiler version 10.1.014); if I "unsetenv NLSPATH", then the complaints "Message Catalog System: corrupt file" disappear. Also, other error types do not have symbolic information either (for example "forrtl: severe (408): fort: (8): Attempt to fetch from allocatable variable U when it is not allocated" or "forrtl: error (63): output conversion error, unit -1, file /dev/ttys000")

[nnemec]

Quoting - Ronald W. Green (Intel)

The lack of symbolic information in the -fpe0 traceback is a bug that has been reported. This bug is being worked.

What is the status of this bug report? I experience a similar behavior with ifort 11.0 but I am not sure whether it is connected.

Greetings,
Norbert

[Kevin_D_Intel]

The lack of a traceback and symbolization of that on Mac OS are fully addressed in the 11.1 release. There were twocomponents to this defect as discussed in a later thread (here).