TapeData2 Notes

	13:27
		Replaced division with conditionals using precalculated values.  Now all
			the (short_int) casts can be removed.
	13:42
		Faster success!  But it still hasnt been tested for loss and noise.
	14:02
		Okay, okay.  CRs and LFs should be allowed to be in the file name.  They
			wont be replaced with spaces anymore.
	14:05
		Instead of just one, the first two samples are ignored.
	14:58
		Fixed a bug generated by the last thing I just said.
		Fixed a problem where the adjustment to avoid same next sample cases was
			interpreted as a next byte.
	15:06
		Fixed a bug where a folder was being created for every file.
	15:22
		Nope.  The adjustment thing has not been fixed.  I looked at the small
			differences (3 bytes) between the result and the original with
			ResCompare.  It appears that there where two entries in the table that
			werent updated ($1B and $27).  Fixing it didnt help much.
	15:32
		I made it a little less sensitive to sharp deltas.
	15:36
		Okay, the remaining problem has to do with alignment out of rebuffering.
			The most recent result had only one inconsistent byte and another byte
			added.
	15:47
		Rebuffering is definitely where it screws up.  Now, instead of $"00" at
			address $1899, it is $"030101".
	16:24
		Turns out it was reading past the buffer while the remaining byte count was
			still positive.
	18:45
		After hair pulling fun, I think Ive figured out what the problem was.
			- how the while loops were constructed.  Once it comes to the end of
			the buffer, it may bail before the part to subtract from the remaining
			byte count.
	19:02
		And just like magic, it worked.  Now for me to get some f**kin sleep.
	19:42
		Well, no magic really.  It worked on the other two files, but not on the one
			I just tested (SDL 1.2 Zip file).
	May 15
	6:14
		After seven hours of sleep for two days worth of sleep (should be 12 hours),
			I am back on to try and make the damn thing work for once!
	7:29-7:35
		After wasting my time devising a sleep formula, I am beckon two triangle
			Mexican damping workforce!
	1PM
		After playing some DOOM II (instead of doing this), I am backne for a
			cracked-knee.  Meaning only exists through interpretation.
	13:11
		Fixed a mistake where temporary was used as a parameter (it should
			have not been that way), and, of course, temporary was used inside
			the function that was called with that parameter!  The name was being
			screwed up once it came to displaying the completion text.
	13:29
		After finding out that the key problem was that it was scanning for the
			next sample chunk at the end of the buffer (before checking the number
			of remaining bytes), I used j, the out-byte position value, to
			store a negative value in this case.  So after adding some code to catch
			the negative j after the next block was read, it goes on and scans for
			the next sample chunk in the same way.  I fixed a mistake regarding it
			not updating the remaining byte count.
	Before 13:57
		Success!
		ResCompare cant seem to handle files without a resource fork.  Or maybe the
			app. has gone corrupt, er something.  Theres no difference between the
			result and the original!
		The bug in ResCompare seems to be larger than that.  After adding a resource
			fork to both files, and even after adding a TEXT resource to both,
			ResCompare still screws up and leaves the blank window when you try to
			compare the files.  Okay, in using HexEdit instead, it could find no
			differences.
	15:00
		Modified the table to have more visible seperation between sample chunks.