rockbox/apps/codecs/dumb/release.txt

/*  _______         ____    __         ___    ___
 * \    _  \       \    /  \  /       \   \  /   /       '   '  '
 *  |  | \  \       |  |    ||         |   \/   |         .      .
 *  |  |  |  |      |  |    ||         ||\  /|  |
 *  |  |  |  |      |  |    ||         || \/ |  |         '  '  '
 *  |  |  |  |      |  |    ||         ||    |  |         .      .
 *  |  |_/  /        \  \__//          ||    |  |
 * /_______/ynamic    \____/niversal  /__\  /____\usic   /|  .  . ibliotheque
 *                                                      /  \
 *                                                     / .  \
 * release.txt - Release notes for DUMB.              / / \  \
 *                                                   | <  /   \_
 *                                                   |  \/ /\   /
 *                                                    \_  /  > /
 *                                                      | \ / /
 *                                                      |  ' /
 *                                                       \__/
 */


******************************************
*** DUMB v0.9.2, released 2 April 2003 ***
******************************************

Yes, there really has been a release. This is not a day-late April fools'
joke.

DUMB's full name has changed! The old "Dedicated Universal Music
Bastardisation" was rather silly, and not much more than a forced attempt at
finding words beginning with D, U, M and B. I spent weeks and weeks browsing
dictionaries and hopelessly asking others for bright ideas, until the
brilliant Chris "Kitty Cat" Robinson came up with "Dynamic". I decided to
keep the U as Universal, since a DUH struct can hold digital music in any
format. Now all that remained was the B, but it didn't take me long to come
up with Bibliotheque, which, despite looking French, is indeed considered an
English word by Oxford English Dictionary Online, to which my university has
a subscription. So there you have it - the name now makes sense.

The two most significant additions to the project would have to be the new
thread safety (with an important restriction, detailed in docs/dumb.txt), and
the new build system. The silly 'makeall' and 'makecore' scripts are gone. If
you are a GCC user, all you need do now is run 'make' and 'make install', as
for other projects. You don't even have to run a 'fix' script any more! There
are some caveats, which are covered in readme.txt. If you use Microsoft
Visual C++ 6, you no longer need to obtain GCC and GNU Make - there is a
project file just for you.

Huge thanks go to Steve Terry for testing on Windows XP - about five times -
and to lillo for testing on BeOS and Mac OS X. Thanks also to X-G for testing
on a Windows system that has consistently posed problems for DUMB's old
makefiles.

There was a bug whereby al_poll_duh() would sometimes cause the music to
resume playing if you called it after al_pause_duh(). Whether this was DUMB's
fault for misusing Allegro's API, or a bug in Allegro, is unclear, but this
release makes it work.

In one of my projects, I found that my AL_DUH_PLAYER stopped playing when
there were lots of other sound effects. In order to fix this, I programmed
DUMB to set the priority of the stream's voice to 255, the maximum. I also
added al_duh_set_priority(), so you can set the priority yourself if you need
to.

The resampling code has undergone a transformation. The bad news is that the
linear average code is no longer in use. The good news is that where DUMB's
resamplers used to require three extra samples' worth of memory to be
allocated and initialised, it now copes with just the sample data. And it
does a very good job at bouncing off loop points and otherwise hurtling
around the sample. The resampling code is considerably more complicated, but
the code that uses the resamplers is considerably simpler - and if you
noticed a slight click in some bidirectionally looping samples, you'll be
pleased to know that that click is gone!

I have also devoted some effort to optimisation. It seemed hopeless for a
while, but then I actually figured out a way of making it faster AND more
accurate at the same time! DUMB is now quite a bit faster than it was, and it
mixes not with 16-bit precision, but with 24-bit precision. (It used 32-bit
integers all along, but the difference is that it now makes use of 256 times
as much of the integer's range.)

There have been the usual improvements to playback. The last release occurred
rather too soon after I had fixed the XM effect memories; EAx and EBx, fine
volume ramps, had been neglected. These are now handled properly.

In previous versions of DUMB, muted channels in IT were actually played with
surround sound panning (where the right-hand channel is inverted). This has
been fixed, so muted channels will really be muted now.

There were also some subtle problems with the way DUMB handled New Note
Actions for IT files. It turned out that, in all releases of DUMB so far,
pitch, filter and panning envelopes and sample vibrato were not being
processed for any note that was forced into the background by a new note on
the same channel! This only affected IT files. Not only has this been fixed,
but envelope interpolation is much more accurate. Long trailing envelope-
driven fade-outs sound a lot better now!

Since panning and filter envelopes are more precise, extra fields have been
added to the DUMB_IT_CHANNEL_STATE struct, used by
dumb_it_sr_get_channel_state(). These fields hold the 'decimal' parts of the
pan and filter cut-off. See dumb.txt for details.

Mxx (set channel volume) now correctly only modifies the last note played on
the channel, not any previous notes that have been forced into the background
by New Note Actions, and filter effect processing is now closer to what
Impulse Tracker does.

The XM loader was slightly flawed and could crash on files containing samples
with out-of-range loop points. One such file was given to me. This has been
fixed.

Finally, the legal stuff. Julien Cugniere has been added to the list of
copyright owners. He deserves it, for all the work he did on the XM support!
And the licence has been changed. You are no longer required to include a
link to DUMB in a project that uses DUMB; the reasons for this relaxation are
explained in licence.txt. However, the request is still there ...

As usual, enjoy!


**********************************************
*** DUMB v0.9.1, released 19 December 2002 ***
**********************************************

Hi again! Lots to say this time, so I shall cut right to the chase.

DUMB now supports Impulse Tracker's low-pass resonant filters! Huge thanks go
to Jeffrey Lim, author of Impulse Tracker, for giving me what information he
still had regarding the algorithm; to cut a long story short, modifying
ModPlug Tracker's source code (which is in the Public Domain) led to an
algorithm whose output matched Impulse Tracker's perfectly.

Please note that ModPlug Tracker's filters as they stand do not match Impulse
Tracker's, and I have no interest in supporting ModPlug Tracker's variant
(especially not the integer rounding problems). Please see docs/modplug.txt,
new in this release, for details.

Thanks also go to Fatso Huuskonen for motivating me to add filter support,
and providing me with several great IT files to test it with!

The other important feature added for this release is click removal. Up until
now, DUMB has generated clicks when cutting notes, starting samples in the
middle, and so on. This version of DUMB will remove any such clicks. Note
that DUMB does not use volume ramps to accomplish this; the algorithm will
not take the bite out of the music!

In other news, DUMB now supports sample vibrato for IT files, and instrument
vibrato for XM files. A slight bug in New Note Action handling for IT files
has been fixed; Note Fade will not break the sustain loops of the sample and
envelope, as it did before. Tremor handling (Ixy) had a strange bug in it,
which has been fixed.

Support for XM files has been greatly enhanced. The XM envelope handling new
in the last release contained a huge bug, resulting in notes seeming not to
stop when they should; this has been fixed. Some XM files crashed DUMB, while
others failed to load; these problems have been solved. Effect memories now
work properly for XM and MOD files, to the best of my knowledge. Some other
differences between IT and XM have been accounted for, most notably the
Retrigger Note effects, Rxy and E9x.

DUMB's sound quality and accuracy are not the only areas that have been
enhanced. The API has been expanded, at last. You can now detect when a
module loops, or make it play through just once. You can ask DUMB to inform
you every time it generates some samples; this is useful for visualisation.
For IT files, you can intercept the MIDI messages generated by Zxx macros,
enabling you to synchronise your game with the music to some extent. (There
is no such method for XM, S3M or MOD files yet; sorry. Also note that the
function will be called before you actually hear the sound; I cannot improve
this until DUMB has its own sound drivers, which won't be for a while.) You
can query the current order and row. Finally, operations like changing the
speed and tempo are now possible, and you can query the playback state on
each channel.

Some parts of DUMB's API have been deprecated. Simple programs that use
Allegro will be unaffected, but if you get some compiler warnings or errors,
please review docs/deprec.txt. This file explains why those parts of the API
were deprecated, and tells you how to adapt your code; the changes you need
to make are straightforward. Sorry for the inconvenience.

For various reasons, I have made DUMB's makefiles use different compiler
flags depending on your GCC version (unless you are using MSVC). There is no
elegant way of getting the makefiles to detect when GCC is upgraded. If you
upgrade GCC, you should execute 'make clean' in order to make DUMB detect the
GCC version again. Otherwise you may get some annoying error messages. (It is
wise to do this in any case, so that all the object files are built with the
same GCC version.)

DUMB's example players have been unified into a single player called
'dumbplay'. The player has been enhanced to display messages when the music
loops, and when XM and MOD files freeze (effect F00; more information on this
in docs/howto.txt).

Finally, as noted on DUMB's website, the release notes from the last release
were inaccurate. It has been verified that DUMBOGG v0.5 does still work with
that release, and still works with this release. The esoteric DUMBOGG v0.6
has not been created yet, since DUMBOGG v0.5 still works.

Please scroll down and read through the indented paragraphs in the notes for
the last release; they are relevant for this release too.

That's all folks! Until next time.


*******************************************
*** DUMB v0.9, released 16 October 2002 ***
*******************************************

MOD support is here! DUMB now supports all four of the common module formats.
As usual, there have also been some improvements to the way modules are
played back. Most notably, handling of tone portamento in IT files has been
improved a lot, and XM envelopes are now processed correctly.

The other major change is that DUMB now does a dummy run through each module
on loading. It stores the playback state at thirty-second intervals. It stops
when the module first loops, and then stores the playback time. This results
in a slightly longer load time and a greater memory overhead, but seeking is
faster (to any point before the module first loops) and the length is
calculated! duh_get_length() will return this and is now documented in
docs/howto.txt and docs/dumb.txt.

DUMB's build process has been changed to use 'mingw' wherever it used
'mingw32' before; some directories have been renamed, and the 'fix' command
you had to run for MinGW has been changed from 'fix mingw32' to 'fix mingw'.

Last time, I directed you to scroll down and read the notes from a past
release, but ignore this point, and that point applies to something else, and
so on. Did anyone do so? Well, if you're reading this at all, you probably
did. Nevertheless, this time I shall be much less confusing and restate any
relevant information. So the least you can do is read it!

- If your program ever aborts with exit code 37 while loading an IT file,
  PLEASE LET ME KNOW! The IT file in question has a stereo compressed sample
  in it, and the format is unspecified for this case (Impulse Tracker itself
  doesn't use stereo samples at all). I will need the IT file in question,
  and any information you can give me about how the IT file was created (e.g.
  what program). (If you don't get to see an exit code, let me know anyway.)

- If your program ever outputs a line resembling "Inst 01 Env: 0,64 8,32
  15,48" to stderr while loading an IT file, PLEASE LET ME KNOW! You have an
  old IT file (saved by an Impulse Tracker version older than 2.00), and
  support for such files is STILL untested.

- Not all parts of DUMB's API are documented yet. You will find some
  functions in dumb.h which are not listed in docs/dumb.txt; the reason is
  that these functions still need work and will probably change. If you
  really, really want to use them, talk to me first (IRC EFnet #dumb is a
  good place for this; see readme.txt for details on using IRC). I intend to
  finalise and document the whole of DUMB's API for Version 1.0.

There have been some changes to the naming conventions in DUMB's undocumented
API. DUMBOGG v0.5 will not work with this and subsequent releases of DUMB;
please upgrade to DUMBOGG v0.6. These changes should not break anything in
your own code, since you didn't use those parts of the API, did you ;)

There is still a great deal of work to be done before DUMB's API can be
finalised, and thus it will be a while before DUMB v1.0 comes out. It should
be worth the wait. In the meantime, there will be 0.9.x releases with
additional functionality, improved playback, and possibly support for some
extra file formats.

Finally I should like to offer an apology; there is a strong possibility that
some of DUMB's official API will change in the near future. There will not be
any drastic changes, and the corresponding changes to your source code will
be simple enough. If I didn't make these changes, DUMB's API would start to
become limited, or messy, or both, so it's for the better. I apologise in
advance for this.

Now scroll down and read the notes for the first r... oh wait, we already did
that. I guess that's it then. You can stop reading now.

Right after you've read this.

And this.

Off you go.

Bye.


********************************************
*** DUMB v0.8.1, released 11 August 2002 ***
********************************************

This is a minor release that fixes a few bugs. One of these bugs, however,
was pretty serious. dumb_register_dat_xm() was never coded! It was prototyped
in aldumb.h, so code would compile, but there would be an unresolved symbol
at the linking stage. This has been fixed.

Platforms other than Unix did not have a working 'make veryclean' target;
this has been fixed. In addition, the makefiles now use 'xcopy' instead of
'copy', since on some systems GNU Make seems to have trouble calling commands
built in to the shell.

Contrary to the errata that was on the DUMB website, the makeall.sh and
makecore.sh scripts actually DID install in /usr. This has now been
corrected, and regardless of whether you use these scripts or call make
directly, the files will now be installed to /usr/local by default.

The XM loader used to treat stereo samples as mono samples with the data for
the right channel positioned after the data for the left channel. This
generally resulted in an unwanted echo effect. This has been fixed.

When playing XM files, specifying an invalid instrument would cause an old
note on that channel to come back (roughly speaking). Fast Tracker 2 does not
exhibit this behaviour. This has been fixed.

The GCC makefiles used -mpentium, which is deprecated in gcc 3.x. This was
generating warnings, and has now been fixed.

In XM files, the length of a sample is stored in bytes. DUMB was assuming
that the length of a 16-bit sample would be even. I had two XM files where
this was not the case, and DUMB was unable to load them. This has been fixed.

In order to accommodate the extra part of the version number,
DUMB_REVISION_VERSION has been added. DUMB_VERSION has also been added in
order to facilitate checking if the version of DUMB installed is sufficient.
See docs/dumb.txt for details.

As a last-minute fix, the XM "Break to row" effect is now loaded properly. It
was necessary to convert from binary-coded decimal to hexadecimal (those who
have experience with Fast Tracker 2 will know what I mean). In short, this
means the effect will now work properly when breaking to row 10 or greater.

DUMB v0.8 had faulty release date constants; DUMB_MONTH and DUMB_DAY were
swapped! For this reason, DUMB_DATE should not be compared against any date
in 2002. This note has been added to docs/dumb.txt and also to dumb.h.

Please scroll to the end and read the release notes for the first version,
DUMB v0.7. Most of them apply equally to this release. However, the
non-portable code was rewritten for DUMB v0.8, so that point does not apply.
The point about length not being calculated also applies to XM files.

Enjoy :)


****************************************
*** DUMB v0.8, released 14 June 2002 ***
****************************************

Welcome to the second release of DUMB!

In addition to these notes, please read below the release notes for the
previous version, DUMB v0.7. Most of them apply equally to this release.
However, the non-portable code has been rewritten; DUMB should now port to
big-endian platforms.

The main improvement in this release of DUMB is the support for XM files.
Enormous thanks go to Julien Cugniere for working on this while I had to
revise for my exams!

There was a mistake in the makefiles in the last release. The debugging
Allegro interface library was mistakenly named libaldmbd.a instead of
libaldmd.a, meaning you had to compile with -laldmbd, contrary to what the
docs said. Apologies to everyone who lost sleep trying to work out what was
wrong! The reason for using libaldmd.a is to maintain compatibility with
plain DOS, where filenames are limited to eight characters (plus a three-
letter extension). The makefiles have now been changed to match the
information in the docs, so you may have to alter your project files
accordingly.

The example programs were faulty, and crashed on Windows if they were unable
to load the file. It was also difficult to work out how to exit them (you had
to click the taskbar button that didn't have a window, then press a key).
They have been improved in both these respects.

I have now added a docs/faq.txt file (Frequently Asked Questions), which is
based on problems and misconceptions people have had with the first release.
Please refer to it before contacting me with problems.

Thanks to networm for touching up the Unix makefile and writing the
instructions on using it.

Incidentally, today (Friday 14 June) is the Robinson College May Ball at
Cambridge Uni. God knows why it's called a May Ball if it's in June. I'm not
going myself (72 GBP, and I'd have to wear a suit, ugh), but with all the
noise outside I shall enjoy pumping up the speakers tonight!


****************************************
*** DUMB v0.7, released 2 March 2002 ***
****************************************

This is the first release of DUMB, and parts of the library are not
crystallised. Don't let this put you off! Provided you don't try to use any
features that aren't documented in docs/dumb.txt, the library should be rock
solid and you should be able to upgrade more or less without problems.

Here are some notes on this release:

- There is some non-portable code in this release of DUMB. It is likely that
  the library will fail to load IT files with compressed samples on
  big-endian machines such as the Apple Macintosh.

- If your program ever aborts with exit code 37 while loading an IT file,
  PLEASE LET ME KNOW! The IT file in question has a stereo compressed sample
  in it, and the format is unspecified for this case (Impulse Tracker itself
  doesn't use stereo samples at all). I will need the IT file in question,
  and any information you can give me about how the IT file was created (e.g.
  what program). (If you don't get to see an exit code, let me know anyway.)

- If your program ever outputs a line resembling "Inst 01 Env: 0,64 8,32
  15,48" to stderr while loading an IT file, PLEASE LET ME KNOW! You have an
  old IT file (saved by an Impulse Tracker version older than 2.00), and
  support for such files is untested.

- The length of IT and S3M files is not currently calculated. It is just set
  to ten minutes.