Compiler Wrapper Problem (sstcc and sst++)

[keitat]

New Issue for sst-macro

The compiler wrappers of SSTMacro (sstcc/sst++) fails to process any C program.

Building SSTMacro using gcc-7.4.0 (or gcc-10.2.0) on CentOS.
./configure --prefix=

This is a test code.

int main(){
 return 1;
}

Here is the output.

./bin/sstcc test.c 
In file included from <command-line>:
/home/knteran/SST/sstmacro-nov21/include/sstmac/skeleton.h:53:36: error: expected identifier or '(' before string constant
   53 | #define SST_DEFINE_EXE_NAME extern "C" const char exe_main_name[] = SST_APP_NAME_QUOTED;
      |                                    ^~~
/home/knteran/SST/sstmacro-nov21/include/sstmac/skeleton.h:212:2: note: in expansion of macro 'SST_DEFINE_EXE_NAME'
  212 |  SST_DEFINE_EXE_NAME \
      |  ^~~~~~~~~~~~~~~~~~~
/home/knteran/SST/sstmacro-nov21/include/sstmac/skeleton.h:206:14: note: in expansion of macro 'USER_MAIN'
  206 | #define main USER_MAIN
      |              ^~~~~~~~~
test.c:2:5: note: in expansion of macro 'main'
    2 | int main(){
      |     ^~~~
In file included from <command-line>:
/home/knteran/SST/sstmacro-nov21/include/sstmac/skeleton.h:214:3: error: initializer element is not constant
  214 |   userSkeletonMainInitFxn(SST_APP_NAME_QUOTED, userSkeletonMain); \
      |   ^~~~~~~~~~~~~~~~~~~~~~~
/home/knteran/SST/sstmacro-nov21/include/sstmac/skeleton.h:206:14: note: in expansion of macro 'USER_MAIN'
  206 | #define main USER_MAIN
      |              ^~~~~~~~~
test.c:2:5: note: in expansion of macro 'main'
    2 | int main(){
      |     ^~~~

When using sst++, it complains that -fPIC is missing.

[knteran@klogin1 sstmacro-nov21]$ ./bin/sst++ test.c 
/usr/bin/ld: test.o: relocation R_X86_64_32 against `.text' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: Nonrepresentable section on output
collect2: error: ld returned 1 exit status

[jpkenny]

sstcc has been broken for years if clang autoskeleton support is not built. The magic that does the renaming of main() uses code that is only valid in C++. It's probably not impossible to fix, but it doesn't currently seem worth the significant effort required. So I've just made the wrapper generate a useful error message and exit for this case.

Not including -fPIC all the time in the wrapper appears to be a conscious decision, but I can't think of a meaningful disadvantage to just compiling -fPIC always. AFAIK every example Makefile we have for skeletons includes -fPIC in the flags. I didn't see any issues in testing so I've changed the wrappers to just add -fPIC to everything. It will probably save some confusion for new users.