
setup.s bugs
- There is a doc bug, setup.s does not go in share/setup, it goes in
  share/vmips
- There is an install bug, setup.s does not actually get installed in
  share/vmips
- DONE: Fixed in trunk and 1.4 branch

I think the reason we are seeing the decserial bug now is that
much of the autoconf stuff is broken. Simplify it.
- --enable-debug/--disable-debug doesn't work
- no-strict-aliasing doesn't work
- DONE: This is addressed in trunk and minimally in 1.4 branch.

Release 1.4.1
- DONE: 1.4.1 is released

The rpm build process is broken. Fix it.
- rpmbuild -ta vmips-<VERSION>.tar.gz should work
- DONE: It works now

Maybe we should incorporate the patches from NetBSD pkgsrc. They are
pretty trivial.
- DONE: Merged to mainline and 1.4 branch (after 1.4.1)

Decserial bug that has to do with store_halfword, that only
appears at -g -O2 (automake's default, for some reason), not at -g3.
- The symptom is 0 being stored to TCR repeatedly when you turn on
  SERIAL_DEBUG.  It only happens when you sh to the TCR address, not sw
- There is a test for it (dzhello)
- The bug is fixed in the little-endian configuration.
  But the test does not pass in the big-endian configuration.
- My current hypothesis is that it should be ok to either use sh or sw
  to poke the serial registers, in little-endian mode. So sh should work
  in either endianness.
- The bug is minimally addressed in 1.4 branch by using the old code but with
  -fno-strict-aliasing. We do not claim that 1.4 boots BSD.
- DONE: Fixed in mainline

etestdev and junk.s would make a worthy regression test.
- need to make it minimally configurable (option to turn it on, off by default)
  - this part is done
- DONE: test is done and passes in both endiannesses (endian-test)

I hate the FSF ChangeLog format. Make something better.
- Parse the output of cvs2cl --xml, I think
- This is in progress. See utils/xmlcl.pl
- Now have diff commands to show the actual content of each commit.
- DONE: New format in use.

FSF address needs to be updated almost everywhere.
- 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
- DONE

Do we want to document tlbsetup.S?  Do we want to replace setup.S with
tlbsetup.S? I'm not sure if there is any point to keeping setup.S if 
tlbsetup.S works.
- DONE: Incorporated tlbsetup features into setup.S. Documented tlb
  setup functions. Removed tlbsetup.S.

It would be better if we could verify that we could actually boot BSD all
the way to a user process.
- Should be able to build a ramdisk into the
  kernel by following the instructions here:
    http://wiki.netbsd.org/tutorials/how_to_create_bootable_netbsd_image/
- Can't let the ramdisk image get too big or else we enter a loop of 
  writing read-only memory.
- DONE: This is verified.

------------------------------------------------------------------------------

We know we can boot NetBSD.
- Need to write down how we accomplished this so other people can reproduce
  the result.
- Make a MIPS NetBSD demo ROM and webpage.
- In principle this should use 1.5 so it looks like progress is happening!

We still want to make a MIPS Linux demo ROM.
- In principle, this should be easily reproducible with the scripts we
  have accumulated. Need to make sure they work.
- This should have its own webpage.
- It would be awesome if we could have a movie to post on the website
  for people to watch.
- In principle this should use 1.4.1

The web site is a mess.  Clear away the cruft.
- No mailing list anymore
  - [DONE] Removed refs to mailing list. Can't find the archives.
- No snapshots for a long time
- No cvsweb
- Docs on web site are old
- I bet the cross tools don't work
- A better design would be nice. Put sections on the left instead of
  across the top

It would make sense to have an "Applications" part of the website
where we collect things like:
- MIPS linux demo
- MIPS netbsd demo

doc install should use install-info, shouldn't it?
- Right now we install a .info file, but no dir entry
- automake nominally has support for
  - building and installing texinfo/pdf/html 
  - building and installing man pages that are generated
- So we should investigate mucking out the doc directory and putting
  automake to work there.
- RPM Spec file needs to be fixed to account for whatever we do here
  - Also make sure bsd ports and pkgsrc still work correctly

Upgrade libopcodes_mips to a slightly newer version
- This isn't really that important so we can put it off indefinitely

