SUMMARY:
The build and install system is much improved, and a bug in 2.18 that caused
some applications to fail is fixed.

PER-FILE CHANGES:
(+: New, -: Removed, *: Cleanup, @: Bugfix, =: Sync with kernel)

[Makefile]
+ Add autodetection for modversions. This means that hacks for modversions in
  the C code are no longer necessary.
- Remove most of install stuff. It is handled by an external install script now.

[README]
* Update install instructions

[*.c]
- Remove modversions hacks

[ov511.c]
+ Make i2c_clockdiv a module parameter since no single value seems to work for
  everyone.
@ Fix a warning when compiling without /proc/video support
* Don't need to check if CONFIG_PROC_FS is defined if already checking for
  CONFIG_VIDEO_PROC_FS
+ Enable usb_make_path() with 2.4.20 kernel
* Use dummy inline functions when /proc support is off. This eliminates a ton of
  #ifdefs
* Un-inline most functions. Inlining them just wastes icache.
@ Replace all sleep_on() calls with wait_event(). This should improve SMP
  stability.
* Move sensor clock divider code here, since it depends on lots of variables
  (bridge type, compression, etc...) that the sensor drivers shouldn't have to
  know about.
@ Disable checking of flags and clipcount in VIDIOCSWIN. Some apps broke after
  it was enabled in 2.18.
+ Enable usb_driver.owner with 2.4.20 kernel
@ Don't do EXPORT_NO_SYMBOLS under 2.5.x

[ov511.h]
+ Enable usb_make_path() with 2.4.20 kernel

[ov511.txt]
+ Document i2c_clockdiv param

[All sensor code]
- Remove clock divider setup code (moved to ov511.c)

[ovfx2.c]
- Remove some ov511-specific bits
@ Fix a warning when compiling without /proc/video support
* Don't need to check if CONFIG_PROC_FS is defined if already checking for
  CONFIG_VIDEO_PROC_FS
+ Enable usb_make_path() with 2.4.20 kernel
* Use dummy inline functions when /proc support is off. This eliminates a ton of
  #ifdefs
* Un-inline most functions. Inlining them just wastes icache.
= Support 2.5.51 USB API changes (struct pt_regs in completion
  handler) (2.18)
@ Replace all sleep_on() calls with wait_event(). This should improve SMP
  stability.
* Simplifications to V4L ioctl() (2.18)
@ Return code for invalid ioctl() command should be -ENOTTY (2.18)
+ Enable usb_driver.owner with 2.4.20 kernel
@ Don't do EXPORT_NO_SYMBOLS under 2.5.x