2019/05/26

08:08

In the Boot RAM I am going to make everything actual chunks so it could be freed more easily.

08:45

This looks a bit strange, but:

My initial RAM value seems to be seeded with its own address? Maybe the boot RAM loader is wrong or the boot RAM writer?

08:47

Okay so the first initializer is that 0x00000004 should not be 4 but it should be the address of the next block. So perhaps the block links are incorrect?

08:49

So I also have in the initializer writing 0 (8) + 248 => 248. So 248 should be written to 0x00000004. So a byte is written then the address. And it is read back in the correct order.

08:54

Okay so I have confirmed that 248 is written to the first two addresses.

08:59

Actually scrolling by I see (m=null, a=00000000, v=8) so somewhere in the bootstrap there is a bad write that is not writing in the correct offset position, so it ends up being replaced.

09:01

Definitely saw an error in the static field initializers, it just used the static RAM offset. The values 4 and 8 being written makes sense because those are the values for the offset static fields.

09:02

Now the RAM links run longer, so I think it is working.

15:08

So I need like actual method indexes and offsets, along with initializing the vtable for classes. So do wonder how I will do this... I could either do it in the initializers at static bootram time (faster) or at VM time. But I think it will be easier to do it at bootram time.

17:05

Okay, I do not believe I need an actual vtable for SPECIAL and VIRTUAL invokes. Okay so, here is my logic. Any virtual call will always call the highest virtually replaced method in a class. Now SPECIAL works on constructors and super-classes which are not virtually replaced at all for the most part. So virtual works on the class type of the instance, while special operates on the class type that was specified. It uses the same virtual lookup. It uses the same virtual lookup as normal stuff but it ends at that class instance. Of course this is not the case because. We just want to work on the virtual table of the super class. The only difference being the constructors and such which have slightly different logic. Okay so I got it, maybe. If the class we are targetting is the same... actually forget that. If we are calling an instance initialization method we want to call the exact one. But I am pretty sure I can get away with using the special stuff using only a single table.

18:38

Not sure how I want to do vtable initialization. It is a bit complicated! As there are package bounds for package-private and no bounds for private methods. But this is literally the last thing to do.

20:50

Okay so I need to load the pool for the target method and that has to be done in the instance code. So the vtable also needs to contain the pointer to the constant pool of the class. So I guess I will need to multiply by two and then add one just to get the pool data. That is really the only way I can think of to get both things at once. However, it could just be in another array and I think that is the simplest really. Once I know the class to load the vtable from I can load the constant pool array as well.