07:24
So BasicNamespaceWriter loses EDIS usage and does not use it at all, however
for writing I can always wrap that in a proposed future value changing
operation. So I would need to devise a new class where I can easily specify a
kind of write callback to be performed before a write operation. I can use a
growable/shrinkable byte array (not like ByteDeque) that implements
DataOutput so I can easily write such things with the given interface.
07:46
DataEndianess is a duplicate, I knew this from the start because I did not
want to import cldc-full for some classes when it was potentially not needed
for operations. However, I am considering removing it. However a problem with
this is that ByteOrder is not an enumeration, so that makes it more
difficult to use for example. I cannot just make it an enumeration either.
07:54
I have an infinite recursion where a base project has an optional dependency which depends on the base project.
17:22
I need a means of facilitating an easy to distribute test system where I can
package a single set of projects into a single JAR. For example, I would need
to run tests on the reference implementation on a specific device type. I must
have a script that can do this. In general, I will need a test suite where I
can run CLDC tests and other such things on the environment. Since the
reference implementation I am basing off does not support LIBlets, everything
will need to be exported into a single JAR file, which is quite simple. One
other thing I can do is take my existing class decoder and perform a simple
pass which renames methods and class names using a prefix (method names are
renamed due to Object). This would act in a way as a standalone mini virtual
environment that I could potentially get running on existing VMs and ones
such as J2ME VMs. These would be essentially standalone JARs with a kind of
bridge to run single programs. This would allow me to develop the class
library accordingly. Since I now have a package and building system that is
standalone from the build system, this makes it easier. One considering is
that ProjectList should be able to be assigned a static compiler and
launcher, where if a null argument is passed it will attempt to use the
statically provided one. So if I were to write a Java compiler that could
be the default system. But if I do the translation layer as mentioned before
and have a Java compiler, I could essentially get a standalone environment
setup. Since as it stands right now, since the bulk of my library is just
TODO, I would not be running much when it comes to code anyway. Since I
have been stuck on the JIT for awhile, this could be an alternative where I
can slowly work and refactor it along with implementing the main library.
But the most important thing is being able to run tests on the reference
implementation so I can gauge how things work.
20:53
Seems my global set is not working. I printed the identity hash code for the class and in both instances it is a different number. So class loader must either not be delegating or no parent was set.
22:23
So I have basic singular output, but some dependencies are missing so the JAR might not truly be standalone. The missing dependencies are mostly just optional ones.
22:27
Actually, that is because I do not include any optionals.
22:29
One thing I must handle are services.