FYI to llvmdev. This is a huge win for LLVM. Thanks Reid!
-Chris
--
http://llvm.cs.uiuc.edu/
http://www.nondot.org/~sabre/Projects/
---------- Forwarded message ----------
Date: Sun, 04 Jul 2004 05:26:34 -0700
From: Reid Spencer <reid at x10sys.com>
To: llvm-commits at cs.uiuc.edu
Subject: [llvm-commits] CAUTION: Type != Value
LLVMers,
I have committed one of the major changes on the way to finishing bug
122: Type no longer derives from Value. My changes have all been
committed and they pass 100% of Feature & Regression tests as well as
being able to compile llvm-gcc.
Please be aware of the following changes made by my commits:
* Type no longer derives from Value (yeah!). Consequently, Types are
smaller and hopefully faster. Types no longer have a Uses vector.
* You can't use Value and Type interchangeably any more. When you
mean a Type, use Type*. When you mean a Value, use Value*.
* Value.h no longer #includes <iostream>. I've made the necessary
changes to all the .cpp files but going forward please only
#include <iostream> from a .cpp file
* The bytecode file format changed. Types are now written separately
from values
* There's a new bytecode analysis tool, llvm-bcanalyzer, that will
produce statistics about bytecode files.
I'm pretty certain this doesn't break *anything*, but a change of this
scope isn't without its subtleties. If you suspect something is broken
related to these changes or you need further description/explanation,
please don't hesitate to contact me.
Best Regards,
Reid.
