user: implement mlibc as the libc, finally.
It's finally done.. Signed-off-by: kaguya <vpshinomiya@protonmail.com>
This commit is contained in:
@@ -0,0 +1,41 @@
|
||||
#include <dlfcn.h>
|
||||
#include <stdio.h>
|
||||
#include <assert.h>
|
||||
|
||||
#ifdef USE_HOST_LIBC
|
||||
#define LIBBAR "libnative-bar.so"
|
||||
#define LIBBAZ "libnative-baz.so"
|
||||
#else
|
||||
#define LIBBAR "libbar.so"
|
||||
#define LIBBAZ "libbaz.so"
|
||||
#endif
|
||||
|
||||
int main() {
|
||||
// In this test, we have bar -> foo and baz -> foo (where -> means 'depends on').
|
||||
// All three objects contain a definition of a symbol g. Bar calls into foo to retrieve
|
||||
// what foo thinks g is, but since bar appears earlier in the scope than foo, bar's copy
|
||||
// of g wins.
|
||||
//
|
||||
// Next, we load baz, which is identical to bar. When baz calls into foo to retrieve g,
|
||||
// foo still sees bar's definition of g, so bar's copy of g wins.
|
||||
//
|
||||
// Swapping the load order of bar and baz should therefore change the value of g which
|
||||
// foo sees. This behaviour is why dlmopen exists. If we ever implement that, we should
|
||||
// write a similar test and assert that the calls return different results.
|
||||
|
||||
void *libbar = dlopen(LIBBAR, RTLD_LAZY | RTLD_LOCAL);
|
||||
int (*call_bar)(void) = dlsym(libbar, "call_bar");
|
||||
printf("call_bar: %d\n", call_bar());
|
||||
assert(call_bar() == 1);
|
||||
|
||||
void *libbaz = dlopen(LIBBAZ, RTLD_LAZY | RTLD_LOCAL);
|
||||
int (*call_baz)(void) = dlsym(libbaz, "call_baz");
|
||||
printf("call_baz: %d\n", call_baz());
|
||||
assert(call_baz() == 1);
|
||||
|
||||
|
||||
dlclose(libbar);
|
||||
dlclose(libbaz);
|
||||
|
||||
return 0;
|
||||
}
|
||||
Reference in New Issue
Block a user