dosfstools 4.1 - released 2017-01-24
====================================

A test suite is now included and can be run with "make check". The test suite
requires xxd from the vim distribution to recreate image files from hexdumps and
also adds a dependency on automake 1.11 or later for building from sources other
than the release tarball.

Now the default for mkfs for filesystems smaller than 512 MB is 64 heads / 32
sectors to make the defaults a bit more in line with the old defaults used up to
3.0.28.

Other changes are fixes, some of them user visible, such as setting the sector
size by command line option for mkfs working again after being broken in 4.0.

Fixed a possible fatlabel crash when writing a label to an unlabelled filesystem
and the given label contains printf style format specifiers.

When circular cluster chains were detected and the -t option was not given, fsck
used to truncate the file to zero length while not updating the free cluster
count. This required a second fsck run to clear. Now it always truncates to the
last cluster that begins the loop, as it is supposed to.

The parsing of octal character specifications for filenames in the -u and -d
options of fsck now works. It was broken in some ancient version and would
always fail.


dosfstools 4.0 - released 2016-05-06
====================================

The programs are now portable to non-Linux operating systems. To that end, the
build system has been converted to use autotools. There have been Linux
specifics in a lot of places which have been either eliminated or should have
equivalents so that it should now work in other Unix-like environments. It has
been tested on FreeBSD and OS X.

As part of making it portable all the code that assumed 8 bit major/minor
numbers - and in fact masked out all other bits - has been cleaned up. Now
mkfs.vfat should not misidentify devices anymore and require the -I option to
override. The new device probing uses libudev (if available) to collect more
information.

Fixed data corruption errors in fsck.fat: Writing to the third to last cluster
on FAT12 with an odd number of clusters would corrupt the following cluster. In
mkfs.fat, long existing bugs in bad cluster marking (from scanning or user
supplied bad blocks list) were fixed so that it actually marks the correct
clusters.

The automatic alignment of data clusters that was added in 3.0.8 and broken for
FAT32 starting with 3.0.20 has been reinstated. If you need to create file
systems for finicky devices that have broken FAT implementations use the option
-a to disable alignment.


dosfstools 3.0.28 - released 2015-05-16
=======================================

The major user visible change in this release is that fsck.fat now defaults to
interactive repair mode which previously had to be selected with -r. The
previous default of a read only check mode was confusing to users who had to
repeat a potentially lengthy fsck.fat run with the right option in order to
actually fix their file system. It was also pointless – the interactive repair
mode already won't write anything without asking for confirmation.

mkfs.fat now allows choosing 0xF0 as the media byte which was previously
rejected.

mkfs.fat now supports the --invariant option to facilitate testing mkfs.fat
itself. It will reproducibly generate filesystems without random or time based
differences between them when all else is identical.

Bugs fixed in fsck.fat are a read one byte beyond the end of an allocated array
when checking some FAT12 filesystems, and checking that the first cluster of a
file as specified in the directory entry is not 1. Previously it could attempt
to follow a block chain starting on cluster 1 and segfault when the conditions
are right.


dosfstools 3.0.27 - released 2014-11-12
=======================================

This is a pure bug fix release. The major bugs fixed:

* fatlabel did not recognize long file names and mistook long file name
  segments in the root directory for labels. This caused output of garbage when
  asked to print the label and damage to the root directory (loss of long file
  name after fsck) when used to set the label.

* A fsck.fat check introduced in 3.0.26 triggered use of uninitialized fields
  in the constructed root directory entry, which randomly caused the code
  checking file names to consider the empty "file name" of the root directory
  to be bad:

    $ /sbin/fsck.fat -y bad.img
    fsck.fat 3.0.26 (2014-03-07)
    /
    Bad short file name ().
    Auto-renaming it.
    Renamed to
    bad.img: 14 files, 19388/403266 clusters

* And finally fsck.fat will not print the version string anymore every time the
  -v option is encountered.