// a.c
enum {aValue=1};
int puts(const char*);
static int foo(void){
    puts(__FILE__);
    return aValue;
}

int getA(void){
    return foo();
}

// b.c
enum {bValue=2};
int puts(const char*);
static int foo(void){
    puts(__FILE__);
    return bValue;
}

int getB(void){
    return foo();
}

// c.d
import a;
import b;

void main(){
    int x = a.getA();
    assert(x==a.aValue);
    x = b.getB();
    assert(x==b.bValue);
}

If you compile the C files separately, and run the resulting program the assertions in main pass and you’ll see:

a.c
b.c

and the process exits with a 0 exit code.

However, if you compile it with dmd c.d a.c b.c or a similar command with -i, you’ll see:

a.c
a.c
src/rt/dwarfeh.d:346: uncaught exception reached top of stack
This might happen if you're missing a top level catch in your fiber or signal handler
[email protected](10): Assertion failure

which indicates it is using the definition of foo from a.c.

Discovered this trying to compile sqlite3 with ImportC. Their shell.c example has a different static function named appendText than the one in sqlite3.c so when compiling them together you’d get weird memory corruption.

I also experimented and found that the right function gets called at CTFE. Behavior can be seen both with dmd and current ldc, so I don’t believe the issue is in the backend.