CP/M/CPM_CDROM.iso/jsage/znode3/z3util/cr28src.lbr/CRUNCH28.HZS/CRUNCH28.HIS

Text File | 1991-05-10 | 25.9 KB | 491 lines

                       CRUNCH/UNCR HISTORY


     +--------------------------------------------------------+
     | Previous versions of CRUNCH/UNCR included no history   |
     | file, with the exception of partial histories accom-   |
     | panying the DateStamper (2.3D) versions.  This history |
     | file has been compiled by condensing the contents of   |
     | release notes and other information included in previ- |
     | ous distribution libraries available to me.            |
     |                                       Gene Pizzetta    |
     |                                       March 19, 1991   |
     +--------------------------------------------------------+


Version 2.8 -- May 5, 1991 -- Gene Pizzetta
     Fix one bug and create another!  But this one is very minor:
     if the file was copied (rather than crunched) the date stamp
     was written to the destination file twice.  And if the A
     option was used, the  archive attribute was set twice on the
     source file (the latter dates from version 2.4).  Thanks to
     Howard Goldstein for finding that one.

Version 2.7 -- April 27, 1991 -- Gene Pizzetta
     Implemented Bruce Morgen's suggestion for a ZCPR3-only
     version that would allow UNCR to be smaller than 8K and
     still uncrunch LZH files.  In the ZCPR3-only version the Z80
     CPU test is replaced with a Z-System check.  Also fixed a
     bug: Bruce Morgen reported that CRUNCH with the A option
     failed to set the archive attribute on the source file if
     the destination file was in a different user area.  Putting
     the call to DATSTP after the call to ARCIT fixed that
     problem.  Howard Goldstein reported that if either program
     aborted because of insufficient memory, an arbitrary number
     of "files processed" was displayed.  The file count is
     initialized later in the code, but zero files is now
     displayed in that case.  Minor screen format corrections:
     "[ file empty ]" message was flush left, instead of one
     space in, as it should have been; and the disk change prompt
     was being overwritten by the next file crunched, for lack of
     a carriage return/line feed.  ZMAC/ZML instructions added to
     TEC file.

Version 2.6 -- March 26, 1991 -- Gene Pizzetta
     Failure to initialize TRPFLG for each file sometimes caused
     an extra null code to be inserted near the beginning of the
     crunched file.  This bug in no way affected the validity of
     the crunched file.  Also failed to include an LZH equate in
     CRUNCH.Z80 after adding LZH uncrunching to UNCR.

Version 2.5 -- March 21, 1991 -- Gene Pizzetta
     When running under ZSDOS, now copies the source file date
     stamp to the output file when crunching, uncrunching, or
     just copying.  In addition, the date stamp is stored in the
     header of the crunched file, in the same manner as the 2.3D
     versions.  UNCR gives priority to the embedded date stamp,
     it it exists, for the output file.  In either case the
     current date will be used as the access date.  If the modify
     date is blank (a non-standard condition), it will be
     supplied from the create date.

     New S option toggles between including and excluding system
     (hidden) files.  As distributed, system files are excluded
     by default.  The old O option has been renamed the E option
     (erase existing files), and the old C option has been
     renamed the I option (inspect mode), for compatibility with
     other standard ZCPR3 utilities.  The old options still work,
     though, for those who have become accustomed to them.

     Now handles up to 512 matching files, if an ambiguous
     filename is given.  (The number of files can easily be
     increased, if anybody has need for it.)  Version 2.4 could
     handle 256 and version 2.3D only 127 matching files.

     UNCR now uncrunches LZH-encoded files, using the R. Warren's
     UNLZH.REL module.  It checks the header of LZH files for an
     embedded date stamp, although no LZH cruncher inserts the
     date at this time.  Nevertheless, we're ready if it ever
     happens.

     Screen displays have been compacted somewhat to allow
     display of more filenames on the screen (twice as many in
     quiet mode).  High bits are now filtered when displaying
     filenames so you won't see weird characters.  (For UNCR the
     quiet mode screen looks the same, no matter what kind of
     file is being decompressed.  Each file takes a single lines,
     unless it's a direct copy.)

     Under ZCPR3 the command processor parses the filespecs, so
     all permissible user areas are available.  Under ZCPR33 and
     above, invalid directory specs will generate an error rather
     than defaulting to the current directory.  The usage screen
     will also reflect the actual name of the COM file, even if
     re-executed with the GO command.

     All options are configurable as the default operating mode.
     The usage screen now reflects the current effect of the
     command line options, if they are used.  All configuration,
     including CRUNCH's exclusion list, can be done with ZCNFG.
     The same CFG file is used for both CRUNCH and UNCR, but some
     configuration options are not effective for both programs.
     See the ZCNFG help screens for full details.

     Other changes:  If the original file already has a "Z" as
     the middle character of its filetype (for example, "AZM"),
     CRUNCH will now change only the last character of the
     filetype to a "Z", unless that character is also already a
     "Z".  If either program is aborted with a ^C, the partial
     output file will be closed and erased, so zero-length files
     will no longer be left behind.  Incorporates routines from
     version 2.3D that traps and alters any occurrence of the
     command mode sequence used by Telenet and PC-Pursuit, so
     file transfers will not be aborted on those systems.  CRUNCH
     now recognizes LZH encoded files as already crunched.
     Various messages have been added or altered to be more
     specific or for aesthetic reasons.

     This version is a direct descendant of version 2.4.  Other
     than the TRAPIT routines, none of the version 2.5 code comes
     from the various 2.3D versions, although those versions
     originated the embedded date stamp idea incorporated herein.
     Many, many thanks go to Howard Goldstein for his ideas, his
     help, and his bug finding skills.

Version 2.4 -- September 15, 1987 -- Steven Greenberg
     CRUNCH and UNCR 2.4 process and generate files identical to
     version 2.3 (except embedded revision level byte),
     regardless of mode of operation.  A few very observant
     people noticed that v2.3 could output valid (but different)
     files when running in "quiet" mode.  The output of v2.4
     should be identical to v2.3 running in verbose mode, except
     for the embedded revision level byte.  Some changes were
     made in the implementation of the "core" of the algorithms
     for both CRUNCH and UNCRunch -- in the case of CRUNCH,
     conditionals were removed by splitting into three separate
     loops.  In the case of UNCRunch, an unnecessary chase to the
     end of "virtual links" was eliminated by aborting the search
     as soon as an available reassignments lot is found.  Other
     performance improving changes include less time updating the
     screen and dynamic I/O buffer sizing.  Non-time-critical
     "user-interface" changes (e.g., the "tag mode" code, etc.)
     were coded in as straightforward a manner as possible, with
     little regard to code space minimization and even less to
     speed.  The great majority of the changes are user interface
     related:

     A new option lets you select any number of files from a
     group for processing.  After all the selections have been
     made the files will be processed.  Selections are presented
     and processed in alphabetical order.

     If CRUNCH ever creates a file larger than the original, the
     file will automatically be erased and replaced with a direct
     copy of the original.  If the original is already crunched
     or squeezed, or if the filetype matches a user configurable
     exclusion list, such as .LBR or .ARC, a straight copy
     operation will be substituted.  Thus all specified files
     will be transferred in the most efficient manner,
     facilitating the use of CRUNCH for the creation of .LBR's or
     as a backup utility.  Similarly, UNCR will either uncrunch
     or direct-copy all specified files for full restoration.

     If CRUNCH or UNCR fills an output diskette during a wildcard
     operation, the last (partial) file will be deleted and the
     user will be prompted to change diskettes.  Operation will
     then continue, starting with that last file.

     UNCR and CRUNCH take as many as three or four (respectively)
     command line options, in any combination.  Each option
     corresponds to a mode which can be set to default to an
     user-defined state.  The command line toggle will then flip
     the state back on or off.

     The archive mode toggle, when turned on, will only crunch
     files that have changed since they were last backed up
     (based on the CP/M directory archive bit).  When running in
     this mode each input file will be flagged as archived as
     soon as the crunch operation for that file has completed.
     The "prompt before overwrite" mode toggle allows command
     line control of whether the program stops to warn you each
     time it is about to overwrite a file with the same name.

     UNCR now also unsqueezes as an added convenience!  Usage is
     identical.  UNCR will automatically recognize the file's
     format and use the appropriate algorithm.

     Modest speed improvements have been made through a variety
     of techniques, including more data buffering to reduce disk
     activity.  The extended buffers are dynamically allocated to
     take maximum advantage of currently available memory on your
     system.

     Messages inform you when the programs are crunching,
     uncrunching, unsqueezing or copying and why (e.g., "Already
     crunched" or "Zero length").  Includes the old in, out, and
     compression ratio reports as well as a final figure on the
     total number of files processed.

     Current program constraints limit wildcard operations to a
     maximum of 255 files.  Hitting ^C will entirely abort either
     program immediately.  Although probably of limited
     usefulness, ^S will pause the programs (in verbose mode).
     ^W will then resume.

Version 2.3D3 -- August 25, 1989 -- Howard Goldstein
     This version fixes a bug which resulted in failure to
     recognize a datespec if a destination du was entered.
     Changes in the COMMON routines fix problems concerning the
     correct drive for the !!!TIME&.DAT file when running under a
     non-ZCPR3 system.

Version 2.3D2 -- November 7, 1988 -- Carson Wilson
     Previous versions set the UNCRUNCHed file's Last Access
     stamp to the Last Access stamp of the original file.  This
     version sets the Last Access stamp of UNCRUNCHed files to
     the current time and date if a DateStamper compatible clock
     is available via BDOS function 12 (Return Version).  All
     changes are to this file and are marked "2.3D2" for
     reference.

Version 2.3D1.1 -- August 2, 1988 -- Carson Wilson
     Changes in COMMON.LIB:  Replaced 8080 opcodes with Z80
     opcodes for Z80ASM.  Added code to close !!!TIME&.DAT file
     at label RETCCP:.  This was the only way to ensure that the
     file is always closed and set to R/O.  T&D file is only
     closed if we were writing to it.  Modified CLOSTD and OPENTD
     to preserve the T&D file's original archive bit.

Version 2.3D1 -- July 20, 1988 -- Carson Wilson
     UNCRUNCH version 2.3D did not set the !!!TIME&.DAT file to
     read/only on exit.  UNCRUNCH v. 2.3D1 corrects this error.
     The version name was decided upon to avoid confusion with
     UNCRUNCH version 2.4, which adds several features not
     available in 2.3, but does not support DateStamped files.
     UNCR23D1 is released with the approval of Bridger Mitchell,
     author of UNCR23D.  CRUNCH23D remains unchanged.

Version 2.3D -- March 14, 1987 -- Bridger Mitchell, Plu*Perfect Systems
     CRUNCH and UNCRUNCH support DateStamped files.  CRUNCH
     includes the original file's datestamp data in the header of
     the crunched output file.  UNCRUNCH v2.3D uses these data to
     replace the datestamp of the uncrunched file after it has
     been closed.  This means that the uncrunched file will have
     its original time and date.  CRUNCH also supports an
     additional optional field -- a datespec -- which may be used
     to specify a temporal class of files to crunch, for example,
     all files more recent than a specified date and time.

     CRUNCH now monitors the output for the first two bytes of
     the sequence: 00dh, 040h, 00dh (with hi bits possibly set).
     If detected, it inserts the CRUNCH NULCODE.  This prevents a
     crunched file from triggering Telenet/PC-Pursuit's command-
     mode, which can abort a file transfer.  Code is derived from
     C. B. Falconer's CRN v2.5., 86/12/19.

Version 2.3 -- November 15, 1986 -- Steven Greenberg
     The core of this source code is pretty much unchanged since
     the original conception and refinement of CRUNCH v2.0.
     Though attention was made in selecting an algorithm which
     could be implemented in a relatively efficient manner, and
     some care was taken to keep the 'innermost' loops fairly
     streamlined, there is no doubt room for improvement in terms
     of the optimally efficient implementation.  This simply
     means that future releases may run even faster than the
     current one.

     The programs now can be configured for ZCPR use.  To support
     the Z3 environment descriptor, the patch byte locations have
     been shifted up.  If you are going to be patching these
     bytes yourself, refer to the new PATCH23.DOC, included.
     (Note: while the location of these bytes has changed, their
     function has not).  If you are going to use the install
     program CRINSTAL.COM, just make sure to use v2.3 of that
     program, included.  If you make a mistake and use the wrong
     install with the wrong program, you will simply get a
     "Invalid or Incompatible CRUNCH.COM" or some similar
     message.  Usage of v2.3 is identical to that of v2.1.

     Acknowledgements:  ZCPR3 Consultant: Bruce Morgen.  Also
     thanks to (continued from last release): Keith Peterson, Jon
     Schneider, Jay Sage, Gary Inman, Steve Russel, Terry
     Carroll, George Peace, Pete Zuroff, and many others.

Version 2.2 -- November 3, 1986 -- Steven Greenberg
     This library contains version 2.1 of the CRUNCH and UNCRunch
     data compression utilities.  CRUNCH21.LBR, as originally
     released, contained an installation program CRINSTAL.COM.
     This was created as a last minute after-thought, because the
     number of available "patch bytes" was getting out of hand
     (also past experience has shown that people tend to ignore
     patch bytes, which is understandable).  Unfortunately, the
     last minute nature of the program (and its apparent
     simplicity) allowed it to slip by normal multi-system, pre-
     release testing channels.  While the install program worked
     fine under CP/M 3, it worked only on CP/M 3.  I don't know
     why DRI chose to change the operation of low# system calls
     (namely "6"), but the bottom line is that the program would
     not accept any input as a "Yes" response, including the
     answer to the question "Do you want to continue?" (This made
     the program quite difficult to run).

     Though it has been less than 24 hours since the release of
     CRUNCH21.LBR, their is no recalling in this "business" (I
     use the term loosely).  CRUNCH22.LBR contains a modified
     CRINSTAL v2.2 (which uses good old fashioned BDOS #1 calls).

Version 2.1 -- November 2, 1986 -- Steven Greenberg
     Main change from version 2.0:  Provides full "DU:" support
     for source and destination files.  Drive, User, neither, or
     both in either order will be accepted for either the source
     or destination.  Particularly useful for backup (e.g.,
     "CRUNCH <file> B0:") when working with a hard drive.  The
     display format has been correspondingly updated to always
     show source and destination drive and user codes, whether
     they are specified or not.

     You can now type CRUNCH *.* or UNCR *.* in mixed groups of
     files and get the desired results.  CRUNCH *.* will
     automatically not attempt to crunch files which are already
     crunched (or squeezed!).  This decision is made after
     opening the input file and analyzing the first few bytes,
     not by using the middle letter of the filename extension (so
     .AZM files, etc.  will still be crunched).  This complements
     the UNCR v2.0 feature of only uncrunching crunched files
     when *.* is specified, though that is achieved more
     expediently by filename extension analysis.

     Error recovery:  Put half a crunched file in, get half an
     uncrunched file out!  While this is not a recommended mode
     of operation, it is illustrative of a somewhat more
     intelligent handling of various error conditions.  If errors
     are detected within a crunched file, the maximum amount of
     information accumulated will be written and the file closed.
     Furthermore, almost all errors pertaining to a single file
     will not result in an abort of the entire wildcard operation
     -- the program will continue with the rest of the files.
     Major errors, (e.g., destination disk full), will of course
     abort the whole operation.

     Minor changes were made to provide more useful error
     messages (e.g., "Disk Full" for disk full, and "Output
     error" for other, less common output related errors.  The
     usage message has been significantly expanded:  type CRUNCH
     or UNCR with no filename specified to see it.

     Corrected a one count error on input records displayed when
     uncrunching a version 1 crunched file with the version 2
     program.  Big file fans:  an extra decimal place provided on
     the final input K ---> output K to correctly display file
     sizes even if they exceed 1 megabyte.  Note that all four
     "running displays" are in 4 digit fixed format fields, and
     will loop back to zero after 9999.  This is not too uncommon
     on the "cr" column, and would happen on the input or output
     recs columns if a file size exceeded 1.25 megabyte.  This is
     nothing to be concerned with.  Certain other very technical
     internal corrections have been made -- it is recommended
     that you replace your version 2.0 programs with these 2.1
     versions to maintain utmost reliability under all possible
     circumstances.

     The "running displays" update slightly less often now when
     crunching, as it was determined that a non negligible amount
     of time was being wasted updating the screen.  It may appear
     CRUNCH is running slower, but rest assured it's actually
     running faster.

     More user configurable options, plus an install program!  A
     provision to eliminate the 'Result not smaller than
     original.  Save it anyway?' question has been provided.
     When this situation occurs, it is often on very small files
     -- some people prefer unattended wildcard operation which
     will not stop to prompt for user input, even if at the
     expense of a few extra resulting records.  This, along with
     three other options provided as "patches" in v2.0 (
     "Quiet/Verbose", "Prompt before overwrite", and "Defeat
     multi-sector I/O" [see accompanying TURBODOS.WRN for a full
     description of that one]) can now be quickly changed (in
     CRUNCH and UNCR simultaneously) by running CRINSTAL and
     answering a few self-explanatory questions.  See
     CRINSTAL.DOC for simple instructions on how to fire up this
     installation program.  (Note to more technical users: the
     one byte patches can still be made using a debugger.  A few
     additional patches, which the install program does not
     support, can be made as well; these are probably only of
     interest to more technical users anyway).

     Acknowledgements:  The flexible drive/user command line was
     made possible thru use of Jim Lopushinsky's "PARSEFCB.REL"
     module.  Many thanks to Paul Foote & Richie Holmes who have
     been supporters of CRUNCH since its inception; to Bob Freed
     for technical asides; and to John Stensvaag for CRUNCH.BG2.

Version 2.0 -- September 2, 1986 -- Steven Greenberg
     CRUNCH v2.0 embodies all of the concepts employed in the
     UNIX COMPRESS-ARC512 algorithm, but is additionally enhanced
     by a "metastatic code reassignment" facility.  The code
     reassignment is augmented with a refined incremental
     compression ratio analysis for adaptive reset.  This
     provides additional improvement, especially on files with
     certain structural variations.  (It is ironic to note that
     if the file ARC512.EXE is CRUNCHed, the resulting file is
     nearly 14% smaller than the file produced when that program
     is used to compress itself).  Although short files will
     generally produce the same results as the above mentioned
     utilities, CRUNCH saves a few extra bytes by eliminating
     zero fill between code length changes.  Other improvements
     include:

     Use of multi-sector I/O when running in a CP/M 3.0+
     environment (automatically selected when appropriate).  May
     be permanently deactivated if desired.

     Relaxed restrictions on filenames (e.g., files with a "Z" as
     the middle extension character, such as .AZM files, are not
     a problem).

     Improved wildcard operation -- non-critical errors will
     abort only the current file, not the entire operation.

     "Verbose" and "Quiet" modes of operation.  The former gives
     full running progress reports while compressing, including
     number of input and output records, compression ratio,
     "codes assigned" and "codes reassigned".  Although some of
     this information has limited usefulness, it is amusing to
     watch.  "Quiet" may be preferred when using slow (or
     printing) terminals, and will allow the results of up to 24
     wildcard operations to remain on the console.

     Optional prompt before erasure of pre-existing files.  A
     warning will be issued if an existing file is about to be
     over-written.  This feature can be deactivated if desired.

     Full compatibility with all crunched files.  UNCR 2.0 will
     uncrunch all "crunched" files, regardless of which version
     of CRUNCH was used to create them.  CRUNCH v2.0 will always
     use the improved algorithm to create new files.

     "Confirm" mode of operation.  Used in conjunction with a
     wildcard filespec, this option allows selectively crunching
     or uncrunching a subset of all files matching the spec.  The
     user will be prompted (Y/N) for each file matching file; a
     response of "N" will simply move on to the next file.

     Continuous checking for ^C abort.

     Inclusion of a "NOP" code.  In addition to the normal
     special EOF and RESET codes, the new structure sets aside
     two additional codes as reserved (one "NOP", one "SPARE").
     The "NOP" code can be inserted into the data stream at any
     point and will always be ignored by the uncruncher.  One
     possible use of this would be a "Telenet Trap" -- where the
     CRUNCH program would monitor its output and insert a NOP
     code if it saw that the output would otherwise produce an
     undesired sequence (ref: R. Freed's PCP-WARN.MQG and PCP-
     FIX.MQG), thus producing files guaranteed to be
     transmittable while using PC-PURSUIT.  Although the current
     CRUNCH program does not yet monitor for this, the structure
     is already set up so this (or any other sequence) could be
     inhibited at any time, yet all files would remain
     upward/downward compatible.  Other data compression schemes
     do not have this flexibility.

Version 1.2 -- June 16, 1986 -- Steven Greenberg
     Existing v1.0 and v1.1 versions of these utilities should be
     replaced with v1.2.  A few minor bugs have been corrected,
     including one which could cause a possible file error when
     crunching numeric data files or object code.  In addition,
     these versions are up to 20% faster, take source as well as
     destination drive specs, and are still under 2k each.

     The source has been provided for the first time, due to
     quite a number of requests (please read the copyright
     message).  All source code is crunched, by the way.  The
     programs are standalone assemblies, but two "include" files
     are used due to the large overlap between programs.  CRUNCH
     or UNCRunch can be assembled with M-80 by just making sure
     "INCLUDE1.INC" and "INCLUDE2.INC" are on the same drive/user
     area.

Version 1.1 -- no information.

Version 1.0 -- no information.

Original author:

     Steven Greenberg
     203 S. Van Dien Ave
     Ridgewood, NJ  07450
     Voice: [phone]