2015/09/28

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.

16:07

I seem unable to size my scrollbars properly.

20:14

With NARF, one thing I can do is make a downgrader so I can compile any Java 8 application using a bootstrap from my class library for my J2ME phone. Some features such as reflection and such do not work and lambdas would require some processing too. For reflection to work, I would need pseudo reflection. Basically a giant class which stores the needed data for reflection. Another possibility is to write VM in the VM when reflection is used and use some kind of wrapper magic for it. However for certain class implementations that might not work. Or I could write a mini JVM which supports Java 8 for said devices. When it comes to lack of memory (my phone at least has 7MiB) I can use a swap device like interface via JSR75. However JAR files are ZIPs which are compressed. J2ME lacks compression for some reason. So I would need a VM and a downgrader to support the target. It would be interesting however, I could compile my OS on the device. This bootstrap engine could also be used for bootstrapping a Java 8 compiler and such for systems which lack Java 8 and have some kind of JVM at least. If I retain my bootstrap target to say Java 1.3 level then it could work on J2ME and very limited and incomplete virtual machines so to speak.

20:29

As an alternative due to collisions, I can change references to all classes to have a prefix. So every class would just be a duplicated interface of a different name. So instead of say "java.lang.Object" it would be instead "net.multiphasicapps.bootstrap.java.lang.Object". Could also go for something smaller such as "bs.java.lang.Object". My goal for my standard library at least is to keep native methods to a minimum. Anything which requires stuff such as filesystems will just delegate that to a system call class rather than use natives. Maintaining the bootstrap stuff as a binary would be a bit messy though, so I suppose I would need to keep a bootstrap collection as source code which can be compiled by stuff such as GCJ and what-not.

20:35

I suppose to support cleaner building of my code, I should split my HairballLauncher up a bit. A higher level bootstrap which calls differing code routes which eventually converge. I would need to make the bootstrap able to be compiled on really ancient versions of Java without requiring major things such as reflection and such. Well my build system uses reflection. At least with a split hairball I would be able to have it where I do not need to reflect in JavaCompiler APIs.

20:41

So there will be 3 versions. HairballSecondStage345, HairballSecondStage67, and HairballThirdState. The HairballFirstStage will be quite simple as it will just determine which other stage to call. I do wonder though if I can find myself Java before 1.2, which is where 1.2 got all of the collections and such. Might have to setup a Linux VM for that using an ancient version of Debian.

20:47

Actually 5 will have to be alone, since that has been given overhauls of things, so that leaves 234 together as one. Otherwise I would have to write more crippled code for 5.

21:48

I can call the javac command from Java, then use defineClass to load the compiled class into the current run-time. That way I will not have to worry about executing by calling the Java command. Or just use URLClassLoader.