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.
I did not actually sleep, just played video games.
Before I continue further I am going to make the compiler able to be internationalized, but I will only have the default English language as the only locale I write. I only speak English. For my international files, I originally was thinking about doing something to the lines of having it be "key=value" through the entire file. However that would be slow, so I would much prefer an actual database then use a program to inspect or add strings. However most databases are binary, I would prefer one to be text based in that there are no control characters or binary bytes except for newline. I also want it so keys are instantly seekable based on their hashes. The file will look really ugly text wise but it may work. I'd need some kind of recovery method in case one decides to change the layout of the file so it breaks. I would have to use newlines otherwise diffs would be insane when there are gigantic lines of text. Looks like to reduce code bloat a bit I could make my database accessible via the SQL interfaces, however those are in compact2 and the locationalization stuff is in compact1. I could just create a class which handles my textual database, then just port it to the JDBC system later on.
Thought about a URL based file system, instead of using just files, URLs are used instead. Though I also want to make a tag based filesystem so I would have to write my own API to access such data in there. Actually, a very cool feature I just thought of, is compiling direct and outputting into a filesystem via URLs. I had previously thought of an API and server to access file system data so I can create filesystems from my build system to boot from. Now if say the compiler can directly write into it, that would be a bonus. Although the standard Java compiler lacks support for such a thing I can use a FileManager to enable that ability. For my filesystem drivers I can just use the Java 7 NIO file systems code and I will not even have to write my own ABI for accessing everything.
Well services in Java are brain dead and not exactly modular, you can only define services in JAR files. And those jar files must be in the ext directory. This complicates things.
As for previous for bridging file to URI based stuff I could do that with a custom file manager and just have the compiler take in stuff in the form of "//uri/(actual URI is here)". Then this way, the standard compiler could in fact use the URI based stuff. Personally I am thinking of just saying screw it with Java 5 support and just require 7 and up to build. However that would mean you could not build it on older implementations.
I can actually support older compilers, but I need a more complex setup for shell scripts and such. So will need to write a hairball boot strapper of sorts which can build hairball, which builds everything else. So I need complex scripts that can handle the most broken and partially implemented runtimes, including my own. My own would be older versions and more compact versions. For this I made a new annotation called HbStandardClassDep. This annotation defines standard class libraries that are required in order to compile this. The rough builder will first check if they truly exist.
So this rough builder will make it so that is horrible rather than crippling my build system to support older class libraries. There is a saying that goes something similar to "Do not break your code for a broken compiler".
I am still working on the rough builder (slowly with distractions) and I am already getting hit by problems in GCJ. The problem passing a charset to an InputStreamReader to decode specific bytes it may use.