smaller ones. Comment from the HISTORY file:
> Creating a ComponentFactory; moving Component::CreateComponent
> to ComponentFactory::CreateComponent.
> Reworking how components are created, in preparation for
> getting "templates" (such as the testmips machine) to work.
> A list of Components is now generated by the configure script,
> so that all .h files with a COMPONENT(name) command in
> src/include/components/ (including template machines, etc) will
> be creatable using ComponentFactory::CreateComponent.
> The -E command line option now works :-)
> 20080113 Adding a "quiet mode" to the GXemul class, set to true by the
> -q command line option.
> Adding some more simple commands: continue, list-components.
> 20080118 [0.4.x branch] Applying a WDC patch from Oleksandr Tymoshenko
> (dev_wdc.c and bus_pci.c), fixing the order of 16-bit words
> within 32-bit words in the identify struct, and not clearing
> the error on reads from the error register anymore.
> [0.4.x branch] RELEASE 0.4.6.2
> 20080119 Adding a generic GetAttribute function to components. Planned
> attributes are "template", "stable", "machine", "description",
> and "contributors". Unimplemented attributes result in an empty
> string being returned from GetAttribute.
> Moving the ASCII tree generator from TreeCommand.cc to
> Component::GenerateTreeDump.
> Documentation of available machine templates and all components
> is now automagically generated on a "make documentation".
> (doc/machines.html and doc/components.html.)
> 20080120 [0.4.x branch] Some minor OpenFirmware fixes: Adding a
> "memory" property to "/chosen", adding a dummy claim() service.
> Adding CTRL-W handling (remove current/last word) to the
> CommandInterpreter.
> The machines.html and components.html documentation pages now
> point to individual doc/machine_XXX.html and component_XXX.html
> pages, if they exist.
> 20080122 Beginning to rewrite the component state variable
> implementation a bit, in preparation for serialization/
> deserialization, undo/redo, and "run backwards" with data
> tightly tied to components.
> 20080123 Continuing on the state variable stuff. Both strings and
> integers (8-bit through 64-bit, signed and unsigned) can be
> used as state variables so far, but more unit tests are needed.
> Component classes can now also be registered runtime. (This is
> needed to enable meaningful unit testing of some things,
> such as cloning of dummy unit test components.)
> 20080124 Moving doc/machine_*.html and component_*.html into subdirs.
> 20080129 Re-adding detection of host arch (ABI) to the configure script.
> 20080203 Continuing on state variables (unit tests etc).
> 20080205 [0.4.x branch] Switching order of DINCLUDE and XINCLUDE in
> src/devices/Makefile (thanks to Ning Zhao for the tip).
> 20080210 Adding -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC to CFLAGS for
> debug builds.
> 20080222 Beginning on the main run loop, GXemul::ExecuteCycles.
> [0.4.x branch] Adding a DEV_DISK_OFFSET_HIGH32 register, for
> reading/writing the top 32 bits of the offset register (useful
> for emulated 32-bit archs), as suggested by Julian Elischer.
> 20080226 Implementing an IRInstruction base class, and some initial
> IR instructions (simple adds, moves, stores).
> 20080227 Continuing on IRInstruction stuff; unit tests to see that
> automatic native code generation kicks in, etc.
> Implementing a simple IRBlockCache, which is just a linear
> memory range with PROT_EXEC, which is reset when it overflows.
> Beginning on a (so far dummy) IRNativeAMD64 code generator.
> Adding a configure option for disabling native code generation
> support completely.
> 20080229 Rethinking how the IR stuff should work: I'll probably use
> simple and traditional interpretation by default (read instr,
> decode, execute), and only let native code generation via some
> form of IR kick in when necessary. Two advantages: it will be
> easier to add new cpu types (because an interpretation switch/
> case block is very easy to understand), and there will be no
> waste of memory (for IR data) for code which is not commonly
> used.
> Removing IRBlock and IRInstruction, but keeping IRBlockCache
> and IRNativeAMD64 for later.
> The IRBlockCache can now be resized during runtime (and it
> expands automatically if necessary).
