but no ioctl is available ;
 * src/css.c: initialize p_disc_key in the case when key decryption fails.

    + Save the cached key as ASCII in hexadecimal form.
    + Overwrite cached files if the data was invalid.

    because they aren't libdvdcss functions.

    + Workaround in CrackTitleKey for strange discs that report read errors
      seemingly at random. Testcase was on Linux 2.4.20, with a region 2 RPC2
      drive, and the region 1 copy of "Chasing Amy".
    + Used macros instead of numerals where appropriate.
    + Fixed spelling here and there.

  <limits.h> automatically, so PATH_MAX was undefined.

  * ALL: removed trailing spaces in files.
  * ALL: moved everything to C99 integer types.

    to cache title keys on disk.

* src/css.c src/ioctl.c, src/libdvdcss.c: changed the work-around to detect
   if the dvd is encrypted on Win2K in non-administrator mode.
   Because we cannot use an ioctl to get the copyright status of the DVD
   we try to get the disc key and if this succeed, we assume the DVD is
   encrypted, otherwise we assume it to be unencrypted.
   I hope this logic is not too much flawed... at least it seems to be working
   with the few DVDs I've got.

    use function pointers instead of doing the if(WIN2K) test.
  * ./src/device.c: if the target is not a drive name such as F:, we open
    it with the standard libc functions, even under Win32.

* Updated Changelog.
* Merged in hh's patch for broken DVD drives/kernel/whatever.

* src/css.c, src/device.c, src/ioctl.[ch]: We don't need to be in administrator mode
anymore to authenticate the drive on Windows NT/2k/XP. As a result any user can now
play a DVD on these OSs :)
* src/ioctl.c: fixed ioctl_ReadTitleKey which wasn't working on Windows NT/2k/XP.
Because of this bug, the disc and key methods for key decryption where not working.

  * ./test/csstest.c: more documentation.
      see http://www.videolan.org/libdvdcss/doc/

    be reset, we read the first sector of the disc instead of closing and
    reopening it.

    falling back to the title method.

   Note: on my drive, this fixes the "ioctl_ReadTitleKey failed" error
         many users have been reporting. Please test!

   Note 2: I could not find any other way to reset the drive after a failed
           ReadTitleKey ioctl than closing and opening the device again. If
           I don't do that, read() fails with an Input/output error after a
           while (but not immediately).

  * ALL: libdvdcss builds with -ansi -pedantic. Yeah I like that :-)
  * ./src/css.c: renamed a few functions so that they make more sense, added
    debug messages here and there, fixed typos and speling.
  * ./src/error.c: moved _dvdcss_error and _dvdcss_debug here.
  * ./src/device.c: moved device reading functions from libdvdcss.c to here.
  * ./src/device.c: errors from dvdcss_read are now properly handled; partial
    reads still aren't though.
  * ./src/libdvdcss.c: default verbosity is now 0, a library should not be
    intrusive by default.
  * ./test/csstest.c: additional error check.
  * ./configure.in: removed the crap boolean_t detection.

  * ./src/css.c: speling fixes.
  * ALL: re-ran bootstrap with more recent autotools.

* updated INSTALL doc for the win32 build.
* added small win32 specific error message that was also in the vlc tree.

  * ./src/dvdcss/dvdcss.h: moved dvdcss.h in a subdirectory so that we can
    include <dvdcss/dvdcss.h> without having to make install.
  * ./test/test.c: removed #ifdef DVDCSS_DIST.

  * ./ChangeLog: updated changelog.
  * ./Makefile.am: added the debian dir to the targets.
  * ./configure.in: switched to a more conventional versioning scheme.
  * ./src/css.c: removed a useless #ifdef.
  * ./src/dvdcss.h: added a versioned symbol to the API.
  * ./src/Makefile.am: use <dvdcss/dvdcss.h> instead of <videolan/dvdcss.h>.
  * ./test/test.c: test program is now up to date.

cracking code is now each in it's own function.  Also moved the code that
looks for a weak block in the VOB to crack the title key from out from
the loop that reads them.  Added another crack method, it's disabled for
now though.  Lowered some limits on the current 'weak' test.  Added more
comments.

compliant to the SFF-8090 standard.  AGID invalidation should now work
allowing us to recover from hung / failed authentications.  Corrected
CSSGetASF, it does not take an AGID argument.  Move several data structures
from the dvdcss handle to local variables in CSSAuth.  Remove CSSAuth as a
externaly visible function in css.c.

  * Initial commit. Hope it'll work.