DISCLAIMER: These notes are from the defunct k8 project which precedes SquirrelJME. The notes for SquirrelJME start on 2016/02/26! The k8 project was effectively a Java SE 8 operating system and as such all of the notes are in the context of that scope. That project is no longer my goal as SquirrelJME is the spiritual successor to it.
Snowed some more. The new hairball boot code looks much cleaner for the most part and is far simpler also.
Seems even with my newer and still even lighter hairball bootstrap code, I still get the issue of the class stuff.
Porting it to Windows was not too troubling. However I suppose the last thing that I must do is just cheat. That is, just load all of the classes in hairball (or at least reference them) and keep them in memory during the compilation process.
I will also have to include all of the dependencies of hairball also because it cannot find the JSON library either.
I believe I know why it occurs AFTER the compiler is called. The compiler might be looking at the system class loader to provide classes for the compiler.
Actually, the JavaCompiler provided by Oracle just uses tons and tons of classes.
I could just use Hairball's class loader as the system class loader which may work to do what I want to do. However, HairballClassLoader cannot use resource bundles becuase there is a recursive invocation involved.
However, that still causes the issue.
I suppose as an alternative which probably might not work is to make a special class loader which loads system based classes itself and potentially delegates them. Basically when a class is requested, it just getResourceAsStream(class) and then just loads the class data even if it would be derived from the main class loader. I could then also just hide the parent. There might be issues, well maybe not, when it comes to finding dependencies for classes. I could always just try it out and see how it works. Then if a class is not found in the system class loader it just then delegates to a default package list with one for hairball.
And it appears my
SystemClassLoader fixes said issues so to speak.
So although my kludge is really really ugly, it may only be needed for bootstrapping hairball on Java 8. I will have to see if I can avoid this issue in bootstrap environments for older virtual machines so it can still build it.
I just have to make sure my kludge also works on Windows/Wine. I shall see about that however. I really hope it does work without issue. There is definitely a drop in speed slightly due to the way I programmed the kludge so that it would work reliably. Sadly this is pretty much the first time I had to do something such as this in Java. And it appears that it works in Wine also.
So now I can continue what I was doing, it has quite literally been a week since I last touched the code I was working on due to this Hairball refactor. I currently also do not have the meta code generation implemented but so far nothing uses it. The PowerPC code I have written no longer uses it. I just now have to go back to rewriting the assembler/disassembler to be much cleaner than how it was before. I still know my notes for it however.
I do wonder if stuff such as Squirrel Quarrel still runs. I shall check this out also and provide any fixes if that is needed.
Appears services are not being added to the bootstrap JARs it seems.
Might be just
findResources() not removing the leading slash, since
everything I use for the most part is absolutely references rather than just
local. This may fix it.
Believe I need to perform my planned speed increase of
making it use soft references for entries and general management of the map
interface rather than just always iterating over every single entry every