- Compiler warning fixes for gnome-vfs and fuse modules. (Anton)
- Change volume mounting (actually device opening) to happen with
O_EXCL bit set so at least on Linux no one can change the device
block size under our feet. (Anton)
- Change volume mounting and mkntfs to set the device block size to the
sector size using BLKBSZSET ioctl (Linux only). This should be
optimal for performance and should fixes the bug of not being able to
create the backup boot sector if the number of sectors on the device
is odd, the sector size is 512 bytes, and the kernel is 2.4. (Anton)
- Enforce cluster size, mft record size, and index record size to be at
least equal to the sector size and verify they are still valid and in
particular display a warning message if the volume will not be
mountable by the kernel driver (it requires mft record size and index
record size to be below or equal to the system page size which we
determine using sysconf()). (Anton)
users to submit it, thus write speed to very fragmented files
dramatically increased, while write speed to low fragmented files
stayed at approximately same level.
today's arrangement has five handlers:
ntfs_log_handler_fprintf (renamed from printf)
ntfs_log_handler_null (does nothing, default)
ntfs_log_handler_stdout (everything to stdout)
ntfs_log_handler_outerr (info to stdout, errors to stderr)
ntfs_log_handler_stderr (everything to stderr)
the colour handler has gone away to be replaced by a flag:
NTFS_LOG_FLAG_COLOUR
Contributed by Erik Sornes. No side-effect for v1.2 NTFS which is still
the default. The implementation needs some minor work: see e.g. the chkdsk
messages and the result of ntfscmp'ing the two images.
currently it's #ifdef'd out.
tree.c - code for handling directory trees
rich.c - a few helpers without a home (this file will go away soon)
Note: ntfsrm isn't in the build any more (you need to ./configure --enable-rich; make extras). When rm is less intrusive, again, I'll put it back in the build.
antona
cha0smaster
flatcap
szaka
uvman