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.
Played video games for many hours, was not planning on it though. Luckily I managed to get to work coding. Now dependency recursion after working on it seems to be working. Now the good thing (sort of) is that using GCJ which is Java 5 causes my current stuff to fail because there is no javax.tools package and thus no classes in it. However when I switch to Java 7 there is one and it works as intended.
The next goal would be to setup the boot strap directory then build hairball and the compiler at the same time. Copying files with Java is not hard at all.
Before I fall asleep I am going to note down that I can have a callback in MetaSearch whenever some dependency is added to the output list. With this I could copy source files and packages as I go along while setting up a bootstrap directory. Although the best would be to enqueue actions so that there is no wasted effort copying files needlessly if it is going to fail anyway.
Now I can support callbacks and I also have a list of dependencies and whether they are soft or not. I will have to make the interfaces for javax.tools in my own classpath now. This is so that things compile on Java 5. Something is off with my soft and hard dependency code. Actually after thinking about it, I know what it is. Essentially for soft dependencies I fall back onto the same FileChecker which was also passed to the soft list, so I just need to remove that. If soft checking fails it will go to the hard one anyway. Actually, it is the getInfo() method in MetaSearch which falls back to the hard dependencies if a soft one is not found, this needs to be changed. And now it is fixed. A thought lingers on my mind whether or not this system will work on a self hosted k8 since the classpath will be providing all the k8 classes. I will have to figure that out when I get to that point. So switching between GCJ and OpenJDK, JavaCompiler is hard in GCJ but soft in OpenJDK. This is the intended effect.
I had this idea, that my standard POSIX utilities (which will be written in Java) can use the Tool interface so that they may be run at run-time. The main entry points can just then invoke the tool itself. In fact, main could be said tool. There is getSourceVersions() however, but I could just return an empty set for that.
I decided on moving the annotations for hairball down to just k8.a instead of k8.s.hb.a. This means instead of 9 base characters, there are only 4 now. Five bytes may be small but it does add up. Also means I have to type less if I refer to it by the full name.
Going to need a Reader and Iterator for code point support so that way I can support internationalization a bit easier. Also means I need a CodePointWriter also. Now a large storm is passing by which means thunder will be quite distracting.
CodePointReader is not going to work, I will have to figure out another way around that. I should perhaps just make a LongReader class use a Reader but is itself not a Reader at all.