tree inheritance:

init-cleanup

driver todo for megamerge:
nuke ddc1 paths
nuke weak functions
i810 kms
mach64 LinearVidMem removal
openbsd pciaccess fd inversion
fix linux int10 ReadBIOS and/or merge int10 and generic
geode/sis/xgi *apVidMem removal
i128/mga BiosBase removal
sis virtualFrom removal
mach64/rendition max[HV]Value removal
cirrus should prefer 16bpp
external RAMDAC code motion for glint/s3/tga
freedreno oslib.h detangle

unify xinerama protocol ~-1kloc
- add screen id token
- add removal api
- add primary flag dealie?

fold up pre-1.2 randr
s/xf86DefaultModes/DMTModes/ -100
GLX for Xnest (works-ish on other branch)
rootless-merge merge
rootless-using xwayland
glxproxy something something, -15kloc if you can delete it...
maybe merge int10/vbe
fix shadow for multiple pixmaps
top-level loader (almost, ported api, needs motion)
loadable glx for all ddxes
un-special-case colormap privates
redo fbbits a la sna
merge the one snafb bugfix
render: FindGlyph -> FindGlyphs
miValidateTree: 'forward' might be a pessimization
RegionOp's overlapFunc is constant
detach dmx from fb
destroywindow should return void
redo de-sdking of opaque.h

PRIVATE_COLORMAP
only reason this is special is xf86HandleColormaps, which the ddx drivers
call _after_ creating default colormap, but which registers a devprivate.
should find some symbol always called before xf86HC but never called in
randr 1.2 drivers?  xf86ValidateModes perhaps?

XRT Must Die
Class: Drawable
Types: Colormap, Damage, GC, Picture, Pixmap, Window

colormaps:
CreateColormap has a special case for the default cmap
uninstall is skipped for server cmaps? because default?
WindowOptRec->colormap is an id not a pointer
ScreenRec->defColormap is an id not a pointer, but, okay?
CopyColormapAndFree is... odd.
AllocColor's CMAPENTRY special case should be moved
ListInstalledColormaps is godawful
not clear why vfb/nest duplicate micmap so much

pixmap/window:
aaaaaaaaaa

---

FakeClientID(0):
weird lifetimes:
    screensaver window
    applewm/winwm event resource
visuals should use separate allocator (fbconfigs too?)
GetNewFontClientID?
hmm: xwin GenerateAuthorization, auth .Add

FakeClientID(client->index):
    dri2 add_frame_event hurrr
    DRI2CreateDrawable2

---

dmx wraps render, fuck.

Create/Free GlyphSet: Annoying because GlyphSets are display-wide, so there's
no obvious place to add a wrap chain. Likewise Add/FreeGlyphs, although there
the problem is that glyphs embed their own per-screen list! Fuck sake.

RenderCompositeGlyphs is correspondingly weird.

Not entirely sure why SetPictureTransform/Filter are wrapped.