CHANGELOG

0.6.5 (3/7/2010):
-----------------

- Added tests to verify ('v') function and to pre-save checks to look for
  partitions that end before they begin or that are too big for their
  disks.

- Fixed a bug that could cause spurious data to appear in a grown partition
  table.

- Added ability to convert some or all partitions to logical partitions in
  GPT-to-MBR conversion. This feature is limited by the fact that at least
  one free sector must exist immediately prior to each logical partition,
  so it won't do much good if partitions are crammed together. It should be
  possible to convert back to MBR any disk that started that way, provided
  no partitions were added or resized when the disk was in GPT form; and
  disks that were partitioned with Apple's Disk Utility or other tools that
  insert unpartitioned space should also be convertible. CAUTION: THE
  LOGICAL PARTITION CREATION FEATURE DOESN'T TRY TO ALIGN PARTITIONS OR
  PARTITION HEADER DATA TO CYLINDER BOUNDARIES! It's conceivable that some
  older OSes or utilities will object to these disks, although Linux, OS X,
  Windows Vista, and Windows 7 all seem happy with them.

- Fixed bug that caused creation of 0-length file if an incorrect device
  filename was typed.

- The gdisk program now prompts for a device filename if it's called with
  no options. This enables gdisk to do something useful if it's launched by
  double-clicking its icon in a GUI environment.

- Added workaround for bug in some versions of MinGW that caused the
  program to garble input sector numbers.

- The Windows version now works on disks with over-512-byte sectors.
  Tested on a magneto-optical (MO) drive with 2048-byte sectors.

- Added -D (--display-alignment) option to sgdisk, to display sector
  alignment value (by default, 1 for sub-800GiB disks and 8 for disks
  over that size).

- Fixed bug in computation of CHS geometries for protective MBR. This is
  non-critical, since most modern utilities ignore the CHS geometries.
  Concerned users can use the 'n' option on the experts' menu to build new
  protective MBRs with the new algorithm, if desired. (Note that GNU
  Parted, at least, gets this wrong, too.)

- Fixed memory-allocation bug when reading GPT disks with partition tables
  with over 128 entries; could cause program to crash on startup.

0.6.4-2 (2/20/2010):
--------------------

Note: Neither of the following changes affects actual program code, so I've
left the version number in the program at 0.6.4.

- Altered Makefile to pass user's compiler and linker environment
  variables through.

- Added #include to gpttext.cc to enable it to compile on the latest
  GCC versions (it was failing on at least some 4.4.x compilers).

0.6.4 (2/19/2010):
-------------------

- Added -m (--gpttombr) option to sgdisk, enabling conversion of GPT
  disks to MBR format, with a limit of four partitions total, and of course
  without overcoming the 2TiB limit.

- Added -h (--hybrid) option to sgdisk, enabling creation of hybrid
  MBRs. Fewer options are available in sgdisk than in gdisk, though,
  in order to keep the user interface manageable.

- Fixed off-by-one bug in specification of partition when using the
  -T (--transform-bsd) option in sgdisk.

- Changed the code to create a new MBR unique disk signature whenever a new
  protective MBR is generated (when doing an MBR-to-GPT conversion, when
  using the 'n' option on the experts' menu, or when using the 'o' option
  on the main menu, for example). Previous versions attempted to preserve
  the existing MBR disk signature in most cases, but this resulted in
  values of 0x00000000 whenever an empty disk was partitioned, and often in
  other cases, too. Better to risk changing this value too often than to
  leave multiple disks with 0x00000000 values, I think.

- Added transpose ('t' on experts' menu in gdisk; or -r or --transpose in
  sgdisk) command to enable fine-tuning partition order without doing a
  full sort.

- Added code to clear the MBR boot loader when doing an MBR-to-GPT
  conversion. (This was already done in full-disk BSD-to-GPT conversions.)
  This is done because I've seen a few problem reports that make me think
  some MBR boot loaders freak out and hang the system when they encounter
  GPT disks, and/or they attempt to load a second-stage boot loader stored
  in what is now GPT territory, causing a system hang. Since MBR boot
  loaders don't work on GPT disks anyhow (even GRUB needs to be
  reinstalled), this new wiping behavior shouldn't cause any problems, and
  may prevent a few.

- Fixed bug in Windows version that prevented saving backup files.

- Fixed bug that caused second and subsequent partition numbers in
  prompts in hybrid MBR conversion procedure to be displayed in
  hexadecimal.

- Fixed very obscure potential bug in hybrid MBR/GPT synchronization when
  deleting partitions; code wasn't matching partition lengths correctly,
  which would only affect partitions that start at the same point but have
  different lengths in MBR vs. GPT.

- Fixed bug in the -E option to sgdisk; it was actually returning the
  last free sector, not the last free sector in the largest free block.

- Fixed bug in -t option to sgdisk; it was corrupting partition type
  codes.

- Fixed minor alignment bug in partition summary list ('p' from any menu)
  when partition sizes are between 1000 and 1024 units.

- Backup restore function ('l' on recovery & transformation menu) now
  accepts both backups generated by GPT fdisk and backups created by a
  direct copy (via dd, etc.) of the MBR, main GPT header, and main GPT
  partition table, in that order. ("dd if=/dev/sda of=backup.gpt bs=512
  count=34" will do this on Linux for a disk with a typical-sized GPT table
  of 128 entries.)

0.6.3 (2/3/2010):
------------------

- Fixed serious data corruption bug on big-endian (PowerPC and similar)
  systems.

- Changed several GPT fdisk Solaris type codes to correct a duplicate

- Corrected error in GPT fdisk type codes for NetBSD LFS and NetBSD RAID;
  they were identical, but I've now changed NetBSD RAID to A906, which
  is unique.

- Added GUID for IBM General Parallel File System (GPFS) partition type
  code. Somewhat arbitrarily set it to use the 7501 number (MBR code 0x75
  is used by IBM PC/IX, so it's at least the right company, by my loose
  numbering rules....).

- Improved GUID generation. Prior versions generated completely random
  numbers for GUIDs. This works, but is technically a violation of the
  spec. Unix versions now employ libuuid to generate GUIDs in a more
  correct way. The Windows version still generates random numbers, though.

- Turned PartTypes class into a derived class of GUIDData, and renamed
  it to PartType.

- Created new GUIDData class, to replace the original GUIDData struct.

0.6.2 (1/29/2010):
------------------

- The change-type ('t' on main menu) option now changes the partition's
  name *IF* the current name is the generic one for the partition type.
  If the current name is not the generic name, it is NOT changed.

- Fixed bug that caused new protective MBR to not be created when the
  MBR was invalid and the GPT was damaged and the user opts to try to
  use the GPT data.

- Enabled default partition type code of 0700 when creating partitions
  or changing their type codes. (Type 0700, Linux/Windows data, is set if
  the user hits the Enter key alone.)

- Fixed bug in sort ('s' on main menu) option that caused partition
  numbers to begin at more than 1 if the original partition list had
  too many empty partitions before the last one defined.

- Improved code to determine which partition table to load in case of
  CRC mismatches between the partition tables and the stored CRC values
  in the headers.

- Compiles using MinGW (http://www.mingw.org) to create a Windows binary.

- Moved all disk I/O functions to the new DiskIO class. This helps with the
  Windows port; it uses diskio-windows.cc for Windows-specific code,
  diskio-unix.cc for the Linux, FreeBSD, and OS X code, and diskio.cc for
  cross-platform disk I/O code.

- Changed BSD disklabel detection code to be more correct (I think).
  This change has no effect on my test disks, but I hope it'll work
  better on disks with sector sizes other than 512 or 2048.

0.6.1 (1/20/2010):
------------------

- Fixed bug that returned incorrect disk size on 32-bit versions of
  FreeBSD.

- Fixed bug that prevented FreeBSD version from working on disk image
  files.

- Fixed bug that caused BSD disklabel conversion to fail.

0.6.0 (1/15/2010):
------------------

- Fixed bug that caused the convert to MBR function to fail.

- Added support for disks with other than 512-byte sectors.

- Created embryonic sgdisk program.

- Fixed bug that caused relative sector numbers entered by users (e.g,
  "+128M") to be misinterpreted as from the start of the range rather than
  from the default value.

0.5.3 (1/4/2010):
-----------------

- Fixed bug in display of GUIDs when compiled with some versions of GCC.

- Eliminated warnings caused by additional checks in latest versions of
  GCC. These warnings were harmless, but to eliminate them I've added
  more error checking on disk I/O.

- Eliminated unnecessary warnings about potential data loss if the program
  was launched with the -l option or if writes aren't possible.

- Added code to set the partition boundary value based on the physical
  sector size. (FindAlignment() function.) This function, however, works
  only on Linux, and then only if the BLKPBSZGET ioctl is defined. This
  ioctl is new in kernel 2.6.32 or thereabouts.

0.5.2 (12/31/2009):
-------------------

- Modified partition creation function to begin partitions on 8-sector
  boundaries by default. This improves performance on the new Western
  Digital Advanced Format drives. The new 'd' and 'l' options on the
  experts' menu display and change, respectively, the boundary size.

- Tweaked code to produce fewer warnings on the latest versions of
  GCC.

0.5.1:
------

- Made some minor edits to the man page.

- Incorporated RPM .spec file changes contributed by Scott Collier
  (boodle11@gmail.com).

- Changed method of locating and loading backup GPT data, to use the
  main header's pointer, if it's valid, rather than seeking to the
  end of the disk.

- Added 'e' option (relocate backup GPT data structures) to the experts'
  menu.

- Fixed bug that prevented recovery of partitions in case of partially
  damaged GPT data (bad main and good backup or bad backup and good
  main header, for instance).

0.5.0:
------

- Added GPT-to-MBR conversion function. It's very limited, but potentially
  useful in some cases.

- Fixed bug that caused incorrect file sizes to be reported on 32-bit
  Linux, thus causing problems when editing partition tables in disk images
  or when loading GPT backup files.

- Fixed bug that caused bogus CRC error reports when loading backup GPT
  data.

- Reorganized menus. There are now three: the main menu, the experts' menu,
  and the recovery & transformation menu. The last of these has most of the
  items that had been on the earlier versions' experts' menu.

- Added ability to re-load the MBR and generate a fresh GPT from it. This
  is normally identical to quitting and re-running the program, but it
  could be handy if, say, the GPT partitions on a hybrid configuration are
  badly messed up; this will enable using the hybridized partitions as the
  starting point for a new GPT setup.

- The program now generates CHS values for hybrid and GPT-to-MBR conversion
  MBRs. For the moment, the assumption is the maximum number of heads and
  sectors per track (255 and 63, respectively), although the bulk of the
  code supports other values -- it'd just be awkward to enter the data in
  the user interface.

- Fixed minor display bug that caused number of sectors on the disk to be
  shown as 0 on large disks when running 32-bit binaries.

- Reverted 0.4.2's zap (destroy GPT) changes, since I don't want to wipe
  out a valid MBR if the user created that MBR over an older GPT without
  first properly wiping out the GPT, and the user now wants to wipe out
  the GPT.

- Reformatted and edited the man page. Aside from edits related to the
  preceding program changes, I've altered the markup slightly and trimmed
  much of the more tutorial information from the man page to better
  conform to typical terse man page style.

0.4.2:
------

- Code cleanup.

- Fixed very small formatting bug in display of hex code when a match isn't
  found when converting from an MBR/gdisk hex code to a GUID type code.

- Added the ability to work on disk image files (raw files for virtual
  machines, backup images, etc.). The program assumes that all such disk
  image files have 512-byte sectors.

- Added verification prompt to 'o' main-menu option to avoid accidental
  erasures of all partitions.

- The "destroy GPT data structures" option ('z' on the experts' menu) now
  also destroys all EFI GPT (0xEE) partitions in the MBR.

- Added an extra warning to the "destroy GPT data structures" option if an APM
  or BSD disklabel was detected on the disk.

- Added a buffer flush after destroying GPT data structures, to get the OS
  to read the new (empty or MBR-only) partition table.

- Fixed bug that allowed entry of nonexistent partition numbers when creating
  a hybrid MBR.

0.4.1:
------

- Code cleanup/re-organization

- Partition creation function ('n' on main menu) now uses the start of the
  largest available chunk of free space rather than the first available
  sector as the default starting sector number. This should enable easier
  partition creation if there are small bits of free space on the disk.

- You can now specify the end point of a partition by using a minus sign,
  in which case the end point is the default value minus the specified
  size. For instance, "-200M" creates a partition that ends 200MiB before
  the default end point.

- You can now specify the start point of a partition by using a plus or
  minus sign, in which case the start point is the specified distance from
  the start (+) or end (-) of free space. This is exactly the same as the
  new rules for entry of the end point, except that the default value is
  set differently.

- Deleting a partition now checks for a matching hybrid MBR partition, and
  if one is found, it's deleted. Any empty space that then surrounds the
  0xEE (EFI GPT) MBR partitions is then added to the nearby 0xEE partition.
  If no non-0xEE partitions are left, a fresh protective MBR is generated.

- Added hybrid MBR consistency check to the verify ('v') option and to
  pre-write checks. If non-0xEE/non-0x00 MBR partitions without
  corresponding GPT partitions are found, the user is warned. This finding
  does NOT prevent writing the partition table, though.

- Added non-destructive write test when opening the device file, in order
  to detect the problem with FreeBSD being unable to write to disks with
  mounted partitions (or other potential problems).

0.4.0:
------

- Added support for BSD disklabels. The program can now convert disks that
  use "raw" disklabels, with the caveat that the first partition will
  almost certainly need to be deleted because it'll overlap the main GPT
  header; and convert disklabels contained within a GPT (or a former MBR,
  converted to GPT) partition. In the latter case, the 'b' main menu option
  is used.

- Added support for compiling on FreeBSD.

- Fixed bug that could cause crashes or incomplete sorts when sorting
  the partition table.

- New partitions, including converted ones, now take on the name of the
  partition type as a default name.

- Reorganized some code; created a separate C++ class for GPT partitions
  (GPTPart), which replaced a struct and enabled moving code from the
  bloated GPTData class into GPTPart.

- Fixed a bug that produced spurious warnings about unknown sector sizes
  when loading a backup file.

0.3.5:
------

Note: This version was not officially publicly released; I wanted to test
the big-endian support while developing 0.4.0.

- Tweaked the disk type identification code to warn users to re-sync their
  hybrid MBRs when one is detected.

- Tweaked MBR-reading code to ignore 0xEE (EFI GPT) partitions. This will
  only have an effect on a poorly partitioned MBR disk that contains an
  inappropriate EFI GPT partition, or when attempting to recover a
  corrupted disk by using the hybrid MBR for data recovery.

- Added big-endian (PowerPC, etc.) support!

- Added code to identify and warn of the presence of an Apple Partition
  Map (APM) on the disk.

- Enabled MBR conversion code to handle multiple logical partitions.

0.3.4:
------

- Fixed bug that enabled (possibly accidental) entry of MBR type codes of
  0x00 in GPTData::MakeHybrid(). The fix also enables entry of default
  type code by pressing the Enter key when prompted. Applied a similar
  fix to the entry of the type code for the second protective partition,
  if one is used.

- Fixed a typo: "sectors" was spelled "sectprs" in one spot!

- Fixed bug that caused default entry for end sector to be refused if an
  initial value using a plus sign (e.g., "+20G") was also refused.

0.3.3:
------

- Gave users control over the way MBR partitions are assigned to slots in a
  hybrid MBR setup; the original method (putting the 0xEE partition after
  the real partitions) works well for non-boot disks, but both GRUB and
  GRUB2 become confused by this type of setup, so it needs changing.

- Changed "blocks" to "sectors" in GPT and MBR table displays.

- Added "Boot" column to MBR table display; shows an asterisk (*) when the
  partition's status is bootable.

0.3.2:
------

- Changed __DARWIN_UNIX03 to __APPLE__ as code to enable MacOS X support.

- Added the ability to create a hybrid MBR ('h' on experts' menu). This was
  motivated by my discovery that Windows 7 remains brain-dead when it comes
  to the ability to boot from a GPT disk, at least on BIOS-based machines.

- Added 'z' option to experts' menu, to destroy GPT data structures and
  exit. The intent is to use this feature to enable subsequent partitioning
  of the disk using fdisk or other GPT-unaware tools. (GNU Parted will wipe
  the GPT data structures itself when you create a new MBR ["msdos
  disklabel," in Parted parlance], so using Parted is another option.)

- Slightly altered the effect of the 'o' command on the main menu. It now
  blanks out the protective MBR, as well as the GPT data.

0.3.1:
------

- Added Mac OS X support, provided as a patch by David Hubbard
  (david.c.hubbard@gmail.com).

- Fixed bug in disksize() function on Mac OS. (Possibly dependent on the
  kernel and/or GCC version.) The disk size, of type uint64_t, was not
  being passed correctly, so I reorganized the function to return it as
  the function's return value rather than as a parameter. This seems to
  work OK on my Mac OS test system and on both 32- and 64-bit Linux
  systems.

- Fixed off-by-one bug in GPTData::FindLastAvailable().

- Fixed bug that caused display of options after a disk-write error.

- Fixed several incorrect MacOS X partition type GUIDs, thanks to Yves
  Blusseau (1otnwmz02@sneakemail.com).

0.3.0:
------

- Changed version number to 0.3.0, reflecting the fact that I've received
  no significant bug reports and so am elevating the program to "beta"
  status. This change also entailed altering the warning the program
  displays when saving partition table changes.

- Fixed minor bug in CHS geometry of the protective MBR's type EE partition
  (was producing 0x000200 as the start value, but should be 0x000100).
  Should be a non-critical bug since the protective MBR partition
  definition is only there to keep MBR-only disk utilities from messing
  with the disk.

- Added ability to enter GUIDs as single massive strings rather than in
  chunks.

0.2.2:
------

- Added #include directives required to compile the program using GCC
  4.4.0.

0.2.1:
------

- Fixed partition numbering problem in reports of partition overlaps in
  verification function.

- Fixed 1-sector partition size problem when creating new partitions
  (partitions were 1 sector too big when using the +size option).

- Changed BytesToSI() to display values in bytes without decimal points
  (e.g., "512 bytes" rather than "512.0 bytes").

- Added GPTData class member functions to retrieve GPT data structure
  locations on disk; used in my internal-use-only GPT-wiping program.

- Eliminated the "a reboot is recommended" notice after writing the
  partition table.

- Added notice after sorting the partition table to the effect that
  editing /etc/fstab and/or the boot loader files may be required.

- Fixed bug in MBR-reading function that caused 0x0f (Win95 LBA) and 0x85
  (Linux extended) extended partitions to not be read.

- Fixed bug in GetLastSector() (in support.cc) that would have prevented
  correct user entry of over-32-bit sector numbers on 32-bit systems.

- Made some changes/corrections to the partition type list in
  parttypes.cc. Most of these were based on newly-discovered MBR type
  codes for Apple (Mac OS X) filesystems.

- General code cleanup (setting explicit casts, etc.)

0.2.0:
------

- Initial semi-public release