I have found a very strange bug. Using this simple Structure (and no associated C code, because no C calls are made):
public static class Bug extends Structure {
public volatile long num;
protected List<String> getFieldOrder() {
return Arrays.asList(new String[] { "num" } );
}
};And this Java code:
Bug b = new Bug();
b.num = 0xABFFEEEE;
b.writeField("num");
Log.d(TAG, "bug pre: " + b);
b.hashCode();
Log.d(TAG, "bug post: " + b);After b.hashCode() is called, the memory contents are 0:
bug pre: Mylib$Bug(auto-allocated@0xb50016f8 (8 bytes)) {
long num@0=ffffffffabffeeee
}
memory dump
[eeeeffab]
[ffffffff]
bug post: Mylib$Bug(auto-allocated@0xb50016f8 (8 bytes)) {
long num@0=ffffffffabffeeee
}
memory dump
[00000000]
[00000000]Notice the last memory dump, all the bytes are zero.
The CPU abi here is armeabi-v7a (Android) and the JNA is built from the repo at 9f094c2. Does not happen if num is not volatile.
I have found a very strange bug. Using this simple Structure (and no associated C code, because no C calls are made):
And this Java code:
After
b.hashCode()is called, the memory contents are 0:Notice the last memory dump, all the bytes are zero.
The CPU abi here is
armeabi-v7a(Android) and the JNA is built from the repo at 9f094c2. Does not happen ifnumis notvolatile.