I have a Java method that wish to be called from some Native method:
public void onDownloadProgress(int current, int total) {
}
and I am trying to call the above Java method from a native method:
int current = ...
int total = ...
jni_callVoidMethod(
env,
jDownloadCallback,
"onDownloadProgress",
"(II)V",
current,
total
);
jni_callVoidMethod is a helper method and its implementation is:
void jni_callVoidMethod(JNIEnv *env, jobject receiver, const char *methodName, const char *contract, ...) {
jclass clazz = env->GetObjectClass(receiver);
if (NULL != clazz) {
jmethodID method = env->GetMethodID(
clazz,
methodName,
contract
);
if (NULL != method) {
BUILD_VARARGS()
env->CallVoidMethodV(
receiver,
method,
va_args
);
if (env->ExceptionCheck()) { // prints out the exception if one is present
env->ExceptionDescribe();
env->ExceptionClear();
}
va_end(va_args);
}
env->DeleteLocalRef(clazz);
}
}
However, in the Java method I am getting some very weird int values. For example, 3840120912. I am wondering can you directly send a C int to a Java primitive int instead of declaring the parameters in the Java method to become Integer type? Declaring those two parameters to be Integer works for me and I am getting the correct value.
Edit: Implementation of the BUILD_VARARGS macro:
#define BUILD_VARARGS() \
va_list va_args; \
va_start(va_args, contract); \
const char *cur = contract; \
while ('\0' != *cur && '(' != *cur) { /* skip to opening paren */ \
cur++; \
} \
while ('\0' != *cur && ')' != *cur) { /* stop at closing paren */ \
switch (*cur) { \
case 'Z': \
va_arg(va_args, int); /* bool (unsigned 8-bit int) */ \
break; \
case 'B': \
va_arg(va_args, int); /* byte (signed 8-bit int) */ \
break; \
case 'C': \
va_arg(va_args, int); /* char (unsigned 16-bit int) */ \
break; \
case 'S': \
va_arg(va_args, int); /* short (signed 16-bit int) */ \
break; \
case 'I': \
va_arg(va_args, long long); /* int (signed 32-bit int) (must be passed in as a long long) */ \
break; \
case 'J': \
va_arg(va_args, long); /* long (signed 64-bit int) */ \
break; \
case 'F': \
va_arg(va_args, double); /* float (32 bits) */ \
break; \
case 'D': \
va_arg(va_args, double); /* double (64 bits) */ \
break; \
case 'L': \
/* fully-qualified-class */ \
while (';' != *++cur && '\0' != *cur); /* advance to end of class declaration */ \
/* FIXME breaks varargs, seems to not be needed va_arg(va_args, jobject); */ \
break; \
case '[': \
/* TODO type type[] */ \
case '(': \
/* TODO ( arg-types ) ret-type method type */ \
default: \
break; \
} \
cur++; \
}
That explains everything! Your calls to va_arg consume the arguments and when CallVoidMethodV gets to it, it is reading from somewhere else in the stack. From the va_arg manual:
The va_arg() macro expands to an expression that has the type and value of the next argument in the call. The parameter ap is the va_list ap initialized by va_start().
Each call to va_arg() modifies ap so that the next call returns the next argument.
Instead, you should create the va_list and immediately hand it off to CallVoidMethodV:
va_list va_args;
va_start(va_args, contract);
env->CallVoidMethodV(receiver, method, va_args);
va_end(va_args);