Here is the promised change to buffer fields.

I need volunteers to complete this patch; or at least the ok on
proceeding to temporarily break the build.  See below for details.


In order to implement thread-specific let-binding of buffer-local
variables, we must introduce an indirection into the C code, so that
lookups of these variables see the thread-specific value, if any.

This patch enables this by adding an accessor macro for each field of
struct buffer.  For a field named `name', the corresponding accessor
is `BUF_NAME'.  All code outside of the GC must then use these
accessor macros.

On the trunk this results in the same code.  On the concurrency
branch, we will redefine these macros to do the needed thread
indirection.


This patch was automatically generated using the attached script.  The
basic idea is simple: rename all the fields in struct buffer, run
`make', and then use the locations in the error messages to decide
where to rewrite.  If you want to reproduce this, you will need a
modified GCC; I can tell you how to get the patches if you want.

Because this script runs the compiler, it only edits locations which
were included in my build.  So, it is probable that some needed
changes were not done.  Fixing any given build failure is simple, but
requires someone to actually see it.

This is where volunteers come in.  We could either proceed by
committing the patch and fixing the problems piecemeal, or by having a
couple people with other systems apply the patch, do a build, and send
me the logs (or patches relative to mine), then repeat.  You would not
need a modified GCC to do this step; I don't expect enough problems to
warrant more automation.


This same technique could be used to extend HIDE_LISP_IMPLEMENTATION
to all lisp types.  Is that worth doing?


The next patch will be applying the same treatment to struct kboard.
Again, it may be a while before I can get to that.


Let me know what you think.  I am not sure what I will do in the absence
of comments.  Maybe check it in!

Tom