v0.07, June 26th 1998:

* ASCII in instructions and data is now supported!
* CMPM is now assembled ok. Previously, it assembled to EOR.
* End of sourcefile is now always recognized. In v0.06 the program needed
  an 'END'-command.
* Operand checking for many modes implemented. For example 'ABS X' would
  have assembled to 'ABS A', but now this is considered an error.
* More checks for conform Nx implemented.
* Support for 'LIST' and 'NOLIST' commands added, but not implemented.
* TTP runs completely in usermode for better behaviour under multitasking
  OSes.
* Now waits decently for a keypress on exit. :)

v0.07a, June 29th 1998:

* Some optimisations.

v0.08, August 21st 1998:

* NORM instruction was assembled incorrectly. Fixed now.
* Checks for illegal parallel-instructions with full 24bit-instructions
  combinations.
* rx/nx pipeline clash checking!
* '-l' option caused a LODfile to be output even if errors had occurred.
  This LOD was mostly corrupt. Fixed now!
* Shows 'usage'-information when executed with false parameters.

v0.09: September 12th 1998:

* Overall 20% faster assembly, through optimisation for common commands.
* DSM command included for cyclic buffers. The current implementation is
  not completely memory efficient, but at least it works.
* Bugs in MOVEP, JCLR/JS(S)ET and MOVE commands all concerning the usage
  of ">" and "<". Removed now.

v0.09a: September 18th 1998:

* Through popular demand (Hi Sentry23! :)) QDSPasm now has an option to
  enable/disable the waiting for a keypress at the end of a session. By
  standard the waiting is OFF. Using 'w' on the commandline enables the
  wait.

v0.10: January 21st 1999:

* Now outputs symbols in LODfiles. This speeds up debugging alot!
* Buggy handling of arithmetic and brackets in statements is now fixed.

v0.10a: February 17th 1999:

* Oh dear. The previous version had a bug in the short address processing.
  This was because of the addition of memorytypes in addresses. Fixed now.
* Yet again a fix in the "MOVEP" instruction handling. This time you were
  forced to use "<<". This isn't needed anymore.

v0.11: May 2nd 1999:
* Again loads of fixes. This time my thanks go out to Thomas/New Beat. He
  mailed a big bugreport and now 3 bugs are fixed:
* Labels defined twice are now seen as an error. This wasn't that serious,
  but it could have caused confusion when the code takes wrong addresses.
* Labels containing registernames (i.e. "r1_label") were handled incorrectly.
  Fixed now!
* The "REP" instruction had problems with accepting immediate data. You needed
  to use "<" in front of the value to overcome this. Now fixed!
* "MOVEP P:ea,io" instruction generated error. Now handled correctly!

v0.11a: September 11th 1999:
* Fixed a very serious bug concerning ENDC, ENDIF. These were not found
  anymore.

v0.12: January 25th 2000:
* The new year brings new features. Finally INCLUDEs can be used. It works
  quite smoothly as well.
* INCBIN can also be used. Not a fast option, since the binary is loaded
  both passes. But there's always the new version in which this could be
  speeded up.
* Better output of symbols in the lodfiles. X/Y Symbols are now output
  correctly and handled flawlessly by Brainstorms DSP-Debug.
* Another bugfix. Yes, the fun never stops. This time it concerned moving
  immediate data from equates into X/Y registers. Bits 16-23 were fucked up.
  Fixed now. Pray that it may never occur again.
* One more bugfix in the bchg/bset/bclr instruction. This instruction isn't
  the most common one, so it was there for quite a while. It always
  assembled into a X-mem operation even if Y-mem was specified!

v0.13: February 5th 2000:
* Some file-operation bugs removed.
* Bug in TCC handling fixed. The comment following this instruction was seen
  as faulty parallel instructions.
* Bug in parenthesis handling removed.

v0.13a: November 12th 2002:
* Bug in equatehandler removed. Bits 31-24 were fucked and this caused a big
  mess.