20070512 bkw:

- Added "detect-tty" script, which attempts to set DEFAULT_TTY_DEV based
on "uname" output. Fallback device is still /dev/ttyS0, but at least the
user is warned if detection fails.

20070510 bkw:

- Portability fixes for AIX: have to set VMIN and VTIME to 1 and 0 in
set_stdin_raw(), then restore the old values on exit. Also, AIX 5.1.0.0
seems to lack the termios constant CRTSCTS, so wrap it in an #ifdef
(you have to set "clocal" on the TTY in SMIT, on AIX). The VMIN/VTIME
fixes don't hurt anything on OSes where they're not needed, and they
will probably be useful for future Solaris/HP-UX/etc ports.

- Got rid of bletcherous int32/uint32/etc typedefs from cas2wav and
wav2cas. Some systems predefine these (AIX does), causing the compile
to break, and they're not necessary anyway on ANSI-ish systems. Since
they're typedefs, I can't #ifdef them...

20070508 bkw:

- Added --long-options. To avoid license incompatibilites, I had to use
a public domain implementation of getopt (from newlib). mkman.pl had to
be made a little smarter, to deal with the --long-options in the help()
output.

- Numeric arguments are now checked, and if they're not actually numeric,
a fatal error happens, instead of silently treating them as zero.

- Minor portability fix for FreeBSD in interactive_usleep()

- Custom baud rates on FreeBSD. Tested on 5.x, with any luck 6.x will
be the same...

20070507 bkw:

- -L mode now attempts to fix truncated EOF and partial blocks.

- Split up giant cassio.c into multiple files. This should help Christos
with his low memory situation when compiling on his ST Falcon. Still need
further cleanups...

- ANSIfied wav2cas.c and cas2wav.c. Not only are they ANSI C, they also
compile as C++ now.

- Split up giant help text string into multiple strings, since some compilers
may not like a single 20 kbyte string...

- Explicitly disable hardware flow control in set_baud_rate()

- Fix a logical bug: init_atarimax_dev() was trying to clear RTS *before*
hardware flow control (RTS/CTS) was disabled, which happens to work on
Linux, but not on win32, and probably not on other UNIXes.

- Restored crosswin target: It actually *works* on a *real* windows machine,
my problems were VMware-related. Still can't compile with minGW though.

- win32: added "Press enter to exit" prompt.

- Get rid of overly-user-friendly code that prepends "/dev/" to the
serial port, since it causes problems for non-UNIX OSes (DOS/Windows
COM1, COM2 ports, and the ST's modem1, modem2 ports). The "0" and "1"
shortcuts for /dev/ttyS0 and /dev/ttyS1 still exist.

- win32: restores README_windows.txt, restored "winzip" target, rename
TODO and Changelog to end in .txt in the .zip file.

20070501 bkw:

- 0.2.0 public release

- Added -a option I promised Steve I'd add, but forgot in 0.2.0

- 0.2.1 public release (sorry Steve!)

20070429 bkw:

- Add a "spinner" during PRWT's longer than 1 second, to give the user
something to stare at intelligently while he waits :)

- Add -X mode, works like -o, but dumps each stage to a separate file.

- Make -U also strip redundant baud records, document useful use of:
cat foo.cas bar.cas baz.cas | cassio -UE new.cas

- Make -E- write to standard output if cassio is reading an image file
from standard input. Hooray for UNIX pipelines!

- Make -R able to write to standard output (and not attempt to re-open
stdout for reading, in that case). Still a minor infelicity in that
"-R -" works, but "-R-" fails (because the "-" is the image filename,
not an argument to "-R")

- Allow user to refer to serial ports by number only. Giving "0" as the
serial port argument is treated as "/dev/ttyS0". Also, if the serial
port doesn't begin with a / or - or . character, "/dev/" is prepended (so
the user can say "ttyS0" to mean "/dev/ttyS0")

20070428 bkw:

- Show start block of each stage (up to 10 of them) in image info.

- Add "s_ify()", so we can intelligently say "1 block" or "2 blocks".

- In -i mode (without -x), suppress bad block warnings after 10 bad
blocks.

- Made -c/-C/-o/-O able to use "-" for standard output.

- Got rid of -C and -O options; added -e option. Instead of -C or -O,
use -ec or -eo.

20070427 bkw:

- Make -R call send_image() with info_only and prescan_only true. This
simulates running "cassio -I" on the new image file.

- Fix -p option, which got broke when I stopped reading the file twice.

- Allow -p n=0 to remove block n from the stream completely. Only works
with data blocks...

- man page now auto-generated from the help. Requires perl on the build
system. Note to self: distribution should include a cassio.1.prebuilt
or the like, in case perl is not installed.

- The -x option now behaves differently with -i: it hexdumps only
blocks with invalid framing bytes or length. With -I, or alone, it
dumps all blocks, as usual.

20070426 bkw:

- Implemented -R (read from serial port) mode. This code was written
"blind": I don't have a working tape drive to test it with. It does work
OK when reading from another copy of cassio over a null modem cable,
but that proves nothing really... the 16550A UART will sh!t the bed if
the tape motor is too fast/slow, or the tape is stretched, or there
are dropouts. Correct way to do this really would be to connect the
SIO data-out pin to something like RI or RTS, then monitor that pin in
a tight loop (can't do with UART's receive-data pin, at least not with
standard serial driver)... or just use wav2cas.

- Added -z (normalize) option, equivalent to -b600 -s260 -M4000 -l12000 -L.
Most useful for -R option, but also -E/-c/-o and normal operation can benefit.

- Added -M (force medium PRWT), works like -s and -l.

- Added -v (compare/diff images) mode, including -x to hexdump differing
blocks.

- hexdump() can now mark bytes with a ! (bangmap), needed so people can
easily see the differing bytes between two very similar block dumps.

- renamed prescan_file() to prescan_image(), and scan_file() to send_image()

- Got rid of unused FILE* parameters to prescan_image() and send_image().

- Fix bug in handle_unknown_block() where the first 4 bytes of cas_data
were printed instead of the 4-byte cas_record_id.

- Added warning for unknown block types in prescan_image().

- Add -Q (really quiet, no warnings) and -P (progress bar) modes. In
either mode, warnings are suppressed, but a warning count is kept.

- On normal exit, if there were any warnings, the total count of
warnings is displayed (this applies even in -P/-Q modes)

- Fix segfault in -E mode (fclose(edit_file) called twice; I don't
understand why it ever worked)

- Change in-place editing option -EE to -E-

- Add -U option (strip unknown/extra records in -E mode)

20070425 bkw:

- Replace all remaining fprintf(stderr) calls with new info() func.

- Fruitless attempt to use win32 API calls with MinGW, to make cassio.exe
compile without the giant cygwin1.dll. I can't tell if it works or not
(my only Windows install is on VMWare, the problem could be with VMWare
and not my code). Reverted. Will try again with a real Windows install (as
soon as I find a box with a real serial port to install it on!). For that
matter, the attempted Cygwin port appeared to have the same problem. For
all I know it works on a real machine, and the MinGW stuff isn't necessary
(though it would avoid an 8MB cygwin1.dll, so it's desirable anyway).

20070424 bkw:

- Minor changes to interactive mode display

- Add sanity-checking before attempting to use the AtariSIO device. This
lets the user know right away if there's a problem, instead of having to
wait for the first PRWT to elapse before reporting an error. Also,
warn if the loaded atarisio module version doesn't match the compile-time
version (but proceed anyway).

- Change a few more fprintf's to warn() and fatal()

- Add mingw/cygwin related comments

- Increase PATH_LEN from 128 to 4096 in cas2wav and wav2cas, so they
don't choke on really long paths.

20070422 bkw:

- Initial attempt to port cassio to win32, with cygwin and a cross gcc.
It compiles, and jali verified that the resulting cassio.exe runs on
his win XP laptop, but nobody's tested it with a real SIO2PC/Atari setup
yet. Annoyingly, cygwin1.dll is required, and it's 8.4 megs (the actual
cassio.exe binary is only 32k!)

- Added MAINTAINER_MODE define, since the fancy fatal() warning()
prototypes are only needed people changing the code, and they can
cause problems for others.

- rearranged Changelog so newer entries are at the top :)

- Added baud-rate changing to AtariSIO mode. Will have to create a new
atarisio tape patch (one-line code change in driver/atarisio.c)

- silence info messages (Initializing edit mode, etc) that were being
emitted in -q mode

- don't print (Fast forwarding over...) for every single block when using
-k or -F; instead print it once when the correct block is reached.

- precalculate time left and bytes sent for each block during prescan,
then use it to display %left and time left during sending.

- Make rew/FF/stop/etc controls work in pause mode. Also, add "P" indicator
when paused, and leave < or > on screen when rew/ff are used. "rewind
to beginning" takes you OUT of pause mode if you were in it; rewind/ff 1
block puts you IN (or leaves you in) pause mode.

20070421 bkw:

- Replace all hard-coded fprintf(stderr, "!!! Fatal... and *** Warning
occurrences with new fatal() and warn(). Change usage() to take an
error message (or NULL for none). stdarg sure is fun...

- Add status indicator "." during PRWT (changes to "W" during writing,
changes to " " after writing). Need this so the user will understand
why his keystrokes don't take effect immediately (can't pause/rewind in
the middle of writing a block)

- Implement rewind 1 block and rewind to beginning commands

20070420 bkw:

- Public release of cassio 0.1.0

- Fix segfault in handle_fuji_block when setting new description (I
overlooked this last night...)

20070419 bkw:

- Dynamically allocate the cas_data array in the cas_blk.
Was allocating 8K of RAM for every record, when most are only 132
bytes long...

- Add cas2cas and its man page

- change help() output, in an effort to make it more manpage-like (and
thus more parseable by help2man). Unfortunatey help2man isn't cooperating,
so I'll probably change it back or make the cassio binary able to generate
its own damn man page...

- when stdin wasn't set to raw mode, don't set it to cooked mode on exit.
This fixes "cassio -I file.cas 2>&1 | less" (less was setting raw mode
before cassio exited; cassio would then exit and set cooked mode, breaking
less's keyboard input!)

- fix broken -S option

- detect text files, and whether or not they have line numbers.

- more sanity checks and warnings in guess_file_type()

- detect DOS binload files and give a warning that they won't be bootable

- some sanity-checking of arguments (create mode and dump/edit modes can't
be active at the same time; -D option gives a warning if used without
either create or edit)

- some code cleanup and rearranging

- man pages for all binaries.

- dynamically allocate cas_blk's as file is read, build a doubly-linked
list of them; file no longer has to be seekable for prescan_file() to
work, and the file isn't getting read twice any more.

20070418 bkw:

- added xex2cas utility. Man, I hate C++, especially old namespace-unaware
code being compiled on modern gcc's.

- Added -L option (trim 133-byte blocks to 132 bytes), to clean up minor
glitches in otherwise-OK .cas files made by wav2cas.

- Finally figured out how to record CSAVE audio from the Atari DIN/RCA
line-out jack and get it to actually work with wav2cas. Documented
in README_examples.txt

- implemented -c/-C (create .cas img from file) options.

20070417 bkw:

- Make -p better: allow user to set PRWT explicitly, and allow up to 1000
blocks to be listed.

- implemented -t (timing adjust), -o/-O (dump data to file)

20070416 bkw:

- Makefile: auto-detect patched AtariSIO

20070415 bkw:

- Sent beta v0.0.4 to SteveS

- Image Editing mode (at least the beginnings of it).

- Support for AtariSIO (requires included patch for AtariSIO)

- Interactive mode, with ff/pause controls. Disable if reading
.cas image from stdin.


20070411 bkw:

- Sent beta v0.0.3 to charliecron

- Added calls to explicitly close/fclose image and serial port instead
of trusting the OS to do it on exit(). Bone-headed mistake on my part
for not doing this.

- Sent beta v0.0.2 to Beetle and charliecron

- Fixed up the "make slackware" and "make slamd" packages. It's now
possible to build the 32-bit slackware package on a 64-bit box, but
not vice versa.

- Defer call to guess_file_type() until after all blocks have been read
in prescan_file(). This allows guess_file_type() to use information
other than just the first data block to help it guess the file type.
Unfortunately, to make this work, I had to rearrange the code, and now
it's uglier...

- If first block PRWT is too short, automatically activate "-d12"
option, unless there's already a -d option in effect

- Uniform *** Warning and !!! Fatal messages

- Added detection for Mac/65 tokenised files... I doubt this will ever
actually get used, but it was just so *easy* :)

- Added detection for LOAD vs. CLOAD images (long IRG vs. short IRG)

- reformat the image information text for easier reading (work in
progress)

for any block that has a bad checksum
- Added checksum validation for 132-byte blocks, with warnings

- Added -x (hex dump) option

- Fixed stupid bug that made % complete always show as 0.0%

20070410 bkw:

- Initial beta, v0.0.1, sent to SteveS

