libhfs/CHANGES

hfsutils - tools for reading and writing Macintosh HFS volumes
Copyright (C) 1996-1998 Robert Leslie

$Id: CHANGES,v 1.14 1998/11/02 22:08:20 rob Exp $

===============================================================================

Version 3.2.6

  * Added casts and made other changes to satisfy fussy signed/unsigned
    compilers.

  * Fixed a problem in which volumes were not marked as "cleanly unmounted,"
    causing unnecessary scavenging to occur when the volume is mounted again.

  * Fixed a problem where scavenging failed to mark all extents from the final
    B*-tree node.

  * Added a simple next-CNID verification to scavenging.

  * Introduced a flags parameter to hfs_format(); new HFS_OPT_2048 format
    option ensures 2048-byte physical-block file boundaries, compatible with
    hybrid ISO 9660 schemes.

  * New HFS_OPT_NOCACHE mount option inhibits use of libhfs' internal block
    cache, useful for RAM file systems or systems/circumstances where the
    cache may decrease performance.

  * New HFS_OPT_ZERO format/mount option forces all allocated blocks to be
    zero-initialized before use, for systems/circumstances in which blocks
    may otherwise contain random data. (This is primarily a security feature
    with a performance cost.)

Version 3.2.5

  * Miscellaneous internal `const' and other minor changes.

Version 3.2.4

  * Fixed a problem with the handling of catalog and extents clump sizes
    (again; first fixed in 1.15b but re-introduced in 3.0b1, alas.)

  * Modified `configure' once again to search for tcl.h/tk.h header files.

  * Changed the OS interface to operate on full block rather than byte
    offsets.

  * Moved the 800K size check from hfs_format() into v_geometry().

Version 3.2.3

  * Fixed a problem with the B*-tree node splitting code. This could have
    caused the libhfs routines to crash under some rare circumstances.

Version 3.2.2

  * Added support for "blessing" the MacOS System Folder by means of a new
    field in the hfsvolent struct. A new option to `hattrib' can be used to
    set this field, as can a new "bless" Tcl volume command.

  * Modified `configure' to use existing Tcl/Tk configuration information
    rather than reconstruct it.

Version 3.2.1

  * Fixed a minor potential problem where BinHex translation might fail to
    recognize the hqx header.

Version 3.2

  * Fixed a problem determining medium sizes under unusual conditions.

  * Modified MDB `drVCSize', `drVBMCSize', and `drCtlCSize' to new
    Sequoia-supporting `drEmbedSigWord' and `drEmbedExtent'.

Version 3.1.1

  * Fixed a problem related to partition locations on large media.

  * Fixed a problem with suid.c on some systems.

  * Fixed a problem with font selection in `xhfs'.

  * Changed copyouts not to append `.txt' to text files if the filename
    already contains a period.

  * Fixed a few other minor problems.

Version 3.1

  * `hmount' and friends now properly handle relative pathnames, even if
    the current directory changes.

  * Enhanced API for managing partitions: hfs_zero() will create a new,
    empty partition map on any medium. hfs_mkpart() will create a new HFS
    partition (if space allows) in the map. hfs_nparts() will return the
    number of HFS partitions currently existing on the medium.

  * hfs_mount() and hfs_format() are now more strict with the partition
    number passed to them; previously this number was ignored if the medium
    was not partitioned. Now the partition _must_ exist, or else 0 can be
    passed to ignore any partitions and select the entire medium.

  * Changed the API for hfs_create(); this routine now returns an open file
    reference upon successful creation.

  * Added bad block sparing to hfs_format(). This routine now requires two
    additional arguments specifying a list of block numbers which are
    defective on the physical medium and should be mapped out of use.

  * Character set translation is now performed as well as end-of-line
    translation in the "text" transfer mode. The routines convert MacOS
    Standard Roman to equivalent (or near-equivalent) Latin-1 (ISO 8859-1)
    sequences and vice versa. It should be noted however that these
    translations are NOT reversible for all characters.

  * Fixed a bug in xhfs.tcl which failed to include the helper `ctime'
    procedure (fixed in hfs.tcl) after it was removed from tclhfs.c.

  * Further librsrc development.

Version 3.0b2

  * Changed the block cache mechanism so that consecutive blocks are read
    and written in single chunks, significantly improving performance.

  * Improved the efficiency of the internal routines for searching B*-trees
    by reducing the number of times the record keys are unpacked from their
    external format.

Version 3.0b1

  * Package is now configured automatically via GNU `configure' script.

  * Numerous internal changes to support automatic configuration and improve
    portability, as well as reduce code size.

  * Bug fix: the alternate MDB was sometimes written to the wrong location
    on non-partitioned media.

  * Bug fix: the Extents Overflow file could become corrupted under some
    circumstances in which a heavily fragmented file is removed.

  * Bug fix: it was possible for the volume allocation routine to become
    caught in an infinite loop.

  * Significant librsrc development. The library can now read and return
    resources to an application, but cannot create or modify them. The
    library can be used with or without libhfs.

  * hfs_mount() now honors the software volume lock bit and makes the volume
    read-only accordingly.

  * Added a new hfs_vsetattr() routine to allow some volume attributes to
    be modified.

  * Added support for partition data blocks that do not start at the
    beginning of a partition.

Version 2.1

  * Further hfsck development.

  * Physical blocks from the volume are now cached by the library. This
    should dramatically improve performance of the programs on some systems.

  * Modified the hfsvolent and hfsdirent structures to include several new
    fields. The hfs_vstat(), hfs_stat(), and hfs_setattr() functions now
    manipulate the new fields. Note that hfsdirent now includes a separate
    union structure for files and directories.

  * Fixed a problem on some systems with `xhfs' that caused it to terminate
    with the message `Error: can't read "data": no such variable'. The
    problem was caused by the inability to parse the system's `df' command
    output.

Version 2.0

  * Renamed hfs_fork() to hfs_setfork(), and added hfs_getfork().

  * Modified all programs to be setuid-aware. Setuid privileges are only
    engaged when opening devices as HFS volumes; this can be useful to limit
    access to and/or help protect the structural integrity of HFS volumes,
    for example by installing the hfsutils programs setgid.  N.B. This
    implementation may not be completely portable, and is isolated to the
    new source file `./suid.c'.

  * First non-beta release.

Version 1.19b

  * Fixed a problem which prevented large volumes from being properly
    formatted by hfs_format(). This should also allow the library to
    correctly manipulate large volumes in certain other cases.

  * Adapted the code to work with Tcl 7.6 and Tk 4.2.

  * Added an HFS globbing interface to Tcl, and extended the `hfs' shell
    to use it.

  * Modified globbing routine to ignore Finder-invisible files.

  * Rewrote `hls' and `hdir'; many UNIX-like options are now available.

  * Updated the general documentation.

  * Continued development of `hfsck'.

Version 1.18b

  * Resolved many signed/unsigned argument passing conflicts throughout the
    code. It should now compile cleanly even on very strict systems.

  * Changed the interface for hfs_read() and hfs_write() to accept a void *
    buffer pointer rather than demanding char *.

  * Made volume file locking optional for systems that don't support it.

  * Included a few other portability fixes for AIX and BeOS.

Version 1.17b

  * Completed the implementation for `hattrib'; HFS file attributes (type,
    creator, invisible/locked) can now be changed from the command line.

  * Fixed an off-by-one error in computing catalog record key lengths.

  * Fixed a problem with backslash-quoted braces during globbing.

  * Improved Makefile handling.

  * Regularized the handling of the volume-unmounted flag.

  * Streamlined the internal and external interface header files.

  * Added a `parid' field to the hfsdirent structure.

  * Updated libhfs documentation.

  * Implemented a better error message reporting format involving pathnames.

Version 1.16b

  * Improved the robustness of the HFS path resolution routine.

  * Fixed hfs_rename() to better detect the validity of destination paths.

  * Changed mount-time scavenging only to mark bits in the volume bitmap,
    not clear them. This reduces the risk of causing further damage to a
    corrupt disk, leaving serious recovery to more advanced tools.

  * Added file locking to hfs_mount() to prevent concurrent processes from
    potentially creating inconsistencies in a volume.

  * Added a new argument to hfs_mount() for specifying mount flags. This now
    allows a volume to be explicitly mounted read-only, read-write, or don't
    care (whichever is available).

  * Improved detection and handling of multiple same-device mounts by a single
    process.

  * Improved error message reporting for all programs.

  * Fixed a problem where hrenaming a volume caused it to become inaccessible
    until it was hmounted again.

Version 1.15b

  * Increased buffer size for native HFS-to-HFS copies to increase
    performance.

  * Fixed a problem verifying the existence of directory threads.

  * Fixed a problem with the handling of catalog and extents clump sizes.

Version 1.14b

  * The command-line programs now support filename globbing; see hfsutils(1)
    for syntax details.

  * File threads are now managed properly when the associated file is deleted,
    moved, or renamed. However, there is no mechanism yet to explicitly create
    or remove threads, or to use them in catalog searches.

  * The volume bitmap is now reconstructed from the catalog and extents files
    when a volume is mounted that was not previously cleanly unmounted.

  * Directories may now be deleted recursively using `xhfs'.

  * BinHex and MacBinary header fork size validation limitations have been
    relaxed. (MacBinary spec suggested 0x007fffff ~= 8MB as maximum length?)

  * Volumes are now flushed automatically by `xhfs' every 30 seconds to
    reduce risk of accidental data loss.

  * Begun work on `hfsck', a program to verify and correct HFS volume
    consistency.

  * Begun work on `librsrc', a library for manipulating the resource forks
    of HFS files.

Version 1.13b

  * Fixed two bad extent record key sorting bugs; these could have caused
    serious problems on certain fragmented volumes.

  * Fixed a problem related to MacOS default file clump size semantics.

  * Implemented recursive directory copies in `xhfs'.

  * Files and directories may now be renamed with a mere change of case.

  * Multiple files may now be moved with `hrename'.

  * A UNIX pathname of "-" in `hcopy' may now be used to mean stdin or stdout
    when used as the source or destination target, respectively.

Version 1.12b

  * Fix to allow mounting of some miscreant partitioned media.

  * Increased copyin/copyout buffer sizes for better performance.

  * Implemented -a for `hcopy' (automatic transfer mode selection.)

  * HFS timestamps are now relative to the current time zone.

Version 1.11b

  * UNIX-to-HFS BinHex transfers have been implemented. All transfer modes
    are now fully implemented.

  * The overall BinHex implementation now uses the stdio library for
    somewhat better performance.

  * New libhfs routines: hfs_setattr(), hfs_fsetattr() for changing file
    and directory attributes (timestamps, type/creator, flags).

  * The copyin routines now update the relevant file information (timestamps,
    flags) for MacBinary and BinHex transfers.

  * The `hvol' command now provides better output.

  * Modified the installation procedure; separate targets now build and
    install the separate components of the package.

  * Fixed a bug in HFS filename sorting which affects the search routines;
    filenames are now sorted precisely according to MacOS idiosyncrasies.

  * Begun work on (but still unimplemented): hattrib, filename globbing.

Version 1.10b

  * Fixed a serious bug in directory record updates; directories with names
    longer than 21 characters could get corrupted.

  * Implemented hfs_rename() and eliminated hfs_move(); functionality for
    the latter has been included in the former.

  * Added an `hrename' command. Minor changes to hcwd.c to cope with
    volume name or cwd path changes.

  * Minor changes to `xhfs'. Files can now be renamed with the interface.

Version 1.9b

  * Modularized the libhfs code.

  * Resolved more portability issues. The code should now work regardless of
    the size or endianness of the host's datatypes.

Version 1.8.1b

  * Fixed some portability bugs.

Version 1.8b

  * More UI enhancements to `xhfs'.

  * UNIX-to-HFS MacBinary II transfers have been implemented; HFS-to-HFS
    copies have been implemented. Only one transfer mode (UNIX-to-HFS BinHex)
    remains unimplemented.

  * Replaced `hcat' with a generic `hcopy'. All transfer modes are supported.
    (But note: The -a option is not yet supported.)

  * Fixed a bug in the MacBinary II encoding CRC algorithm.

  * Tcl hash tables are now used to maintain open files and volumes.

  * Cleaned up some of the Tcl/Tk interface to aid compatibility with
    various versions.

Version 1.7b

  * Many more UI enhancements to `xhfs'.

  * Files can now be copied from UNIX to HFS volumes with `xhfs' using either
    Text or Raw mode translation.

Version 1.6b

  * Fixed a bug in the catalog manipulation routines which would allow files
    and directories to be created at the same level as the root folder.

  * Many UI enhancements to `xhfs'. Some minor enhancements to `hfs'.

  * hfs_islocked() now returns 1 if a volume is locked (read-only).

  * Extended file attributes to recognize invisible files.

Version 1.5b

  * All cases of B*-tree record insertion have been completed.

  * B*-tree record deletion has been implemented. Files and folders may now
    be created and deleted with no limitations (other than the size of the
    volume.)

  * New commands: hrmdir, hdel

  * File writing and truncation have been implemented.

  * The library now maintains its own concept of "current volume". The last
    volume to be mounted is by default current; the new functions
    hfs_getvol() and hfs_setvol() can be used to get and set the current
    volume, respectively. All other functions will now accept a null volume
    pointer to represent the current volume.

  * Absolute pathnames may now refer to (mounted) volumes other than the one
    specified in the call; in this event the specified volume is ignored.

Version 1.4b

  * Updated the installation procedure. The Tcl and Tk programs can now be
    built and installed optionally.

  * Fixed a minor bug which would cause the header node to be read just
    before completing a directory read.

  * Mounting the same volume more than once now returns a reference to the
    same volume, and does the right thing when each reference is
    independently unmounted.

  * Cleaned up some of the Tcl interface to deal with multiply-referenced
    volumes. `xhfs' can now deal with the same volume on both sides.

  * Renamed hfs_cwdid() to hfs_getcwd(), and added hfs_setcwd().

  * hfs_umount() now closes all open files and directories on the volume
    before unmounting it. All external references to the volume and any
    open files/directories will become invalid.

  * Enabled hfs_umountall() which will effectively unmount all volumes,
    invalidating all external volume and file references. This is a
    catch-all that should be called before a program exits. N.B.: Errors
    during volume flushing and unmounting are not reported.

Version 1.3b

  * Enhanced writability: B*-tree file growth has been implemented. New
    directories can be created until the disk gets full. 

  * Added an option to `xhfs' for creating new directories.

Version 1.2b

  * Enhanced writability: B*-tree node splitting has been implemented. New
    directories can be created until the catalog file gets full.

  * Fixed some minor bugs in version 1.1b.

Version 1.1b

  * Full read-only capability.

  * Limited write capability: new directories can be created only until
    a catalog node gets full.

  * New volumes can be created with hfs_format().