Linker error in libmmt.lib

[lklawrie]

error LNK2005: _ceil already defined in libmmt.lib(ceil_stub.obj) MSVCRT.lib(MSVCR90.dll)

Not sure what to do about it?

(Most of the code is fortran, we have a couple of c libraries and are using openmp)\\

[lklawrie]

Caused by having a CEILING intrinsic call in the code? (yes, i commented it out and it is fine).

Compiler bug?

[mecej4]

The line

libmmt.lib(ceil_stub.obj) MSVCRT.lib(MSVCR90.dll)

(shown above) raises a suspicion that some objects were compiled with /MT and some with /MD, or that inconsistent library choices were made.

libmmd.lib goes with msvcrt.lib and /MD

libmmt.lib goes with libcmt.lib and /MT

[lklawrie]

I see that I have a long standing "ignore" for libcmt.lib

not sure about msvcrt.lib

and I guess I recompile the c libraries with consistent switches -- found one that was compiled with /md in release mode.

in the mean time have implemented a work around for the CEILING call though I guess I could try some other ignores.

[lklawrie]

until i can recompile the c libraries, i've gone with all "debug" versions of the libraries. and that seems to work.

[mecej4]

> found one that was compiled with /md in release mode

Nothing wrong with that; /MD simply specifies linking against DLL libraries (e.g., MSVCRT.LIB). The D in /MD does not mean "debug".

If /MT were used instead, that would indicate linking against LIBCMT.LIB.

[lklawrie]

in the c version, it appeared to be a multithreaded DEBUG dll? (that was the /MD)

[Steven_L_Intel1]

/MD is Multithread DLL. If you want the Debug version, that's /MDd.

When you have C code, it is not safe to ignore libraries to paper over a mismatch, as the C compiler actually generates different object code depending on which library choice was made for the project. If it's just Fortran code, then it doesn't matter as much.

[mecej4]

/MT Multithreaded Static Release: LIBCMT.LIB
/MTd Multithreaded Static Debug: LIBCMTD.LIB

/MD Multithreaded Dynamic Release: MSVCRT.LIB, MSVCR*.DLL
/MDd Multithreaded Dynamic Debug: MSVCRTD.LIB, MSVCR*D.DLL

[lklawrie]

Thanks for enlightening me.

It appears I still have to compile these c libraries (someone else compiled them) and make sure they are using the same switches. In the meantime, i went to the debug versions of these and it added considerable to the test suite run.

And I probably threw in the ignore x.lib comment somewhere "right before release" which is when things tend to come together. (and then never investigated it).