[core] “Replace constants” checkbox causes pseudo-code error

[xmhwws]

Issue details

If “Replace constants” is unchecked, then the following picture is shown:

int i13 = 0;
cArr[i13] = (char) i11;

If “Replace constants” is checked, then the following picture is shown:

private static final int i = 210;
int i13 = i;
cArr[i13] = (char) i11;

You can see that the index value, i13, is 0 one time and 210 the next. after actual testing, the value of i13 is 0, so, this is a bug!

Relevant log output or stacktrace

Provide sample and class/method full name

Link: https://drive.google.com/file/d/1n1SDgo5HkFAJ88vY_jfUNCAXD2uE8rt2/view?usp=sharing
class: com.appsflyer.internal.AFi1aSDKSS
Methods: private static String $$c(int i2, byte b, int i3)

Jadx version

latest version: 4644d1d

[skylot]

@xmhwws thank you for report 👍

Turns out that bytecode can set constant init value for static final fields even if they initialized in class init method.
So static final field i was set to 0 by dex field init section -> jadx save it in constant pool for replacement -> replace some 0 in code with that field -> move initialization from class init method into field declaration -> field now set to 210 🤣

To fix that, I commit an additional check to ignore fields which still used in the code. Unfortunately current usage info collected by jadx do not distinguish type of usage (get or set), but anyway any usage is a good indication that the field is not a real constant and not inlined by compiler, so we don't actually need it for constant replacement.