From: nowhere@never.at   
      
   Kerr-Mudd,John wrote:   
   ...   
   >>> I now have something that works, but I haven't "newline"-proofed it.   
   >>> 213 bytes of pure B64 (if you allow crlfs). Quite a struggle to get   
   >>> the fixups working; that and avoiding invalid fixup addresses (e.g.   
   >>> the 5A to 61 gap). If/when I overcome that last problem I'll post it   
   >>> here.   
   >> Perhaps less struggle if you'd had my hex-editable disass/dump-tool :)   
   > Hah it's a heap of equates to keep it "simple".   
      
   > Here's the B64 decoder (after fixups)   
      
   Let me check... and add comments   
      
   > ax=2Bxx   
   wouldn't ax=30xx fit better ?  But I see yet you start at 071.   
   >   
   > ->u3c l3d   
   > ; 3C is 3D-1; 3D is longest code size <64 chars   
   > 1625:003C 3D 2B 30       cmp          ax,302B   
   > 1625:003F 72 33          jb           0074      ;skip all <2B   
   > 1625:0041 3D 3D 30       cmp          ax,303D   
   > 1625:0044 74 32          jz           0078      ;return on 3D   
   > 1625:0046 3D 61 30       cmp          ax,3061   
   > 1625:0049 72 02          jb           004D      ;**2C02   
   > 1625:004B 2C 06          sub          al,06     ;**2C06   
   > 1625:004D 2C 41          sub          al,41   
   > 1625:004F 79 0D          jns          005E      ;*040D gotnum   
   > 1625:0051 04 45          add          al,45   
   > 1625:0053 3D 33 30       cmp          ax,3033   
   > 1625:0056 77 06          ja           005E      ;*1806 gotnum   
   > 1625:0058 18 C0          sbb          al,al     ;*04C0   
   > 1625:005A 04 41          add          al,41   
   > ; no fixups in the 5A-61 gap!   
   > 1625:005C 48             dec          ax   
   > 1625:005D 48             dec          ax   
   gotnum:   
   > 1625:005E 41             inc          cx   
   > 1625:005F 41             inc          cx   
   > 1625:0060 36 98          ss:cbw                 ;*D398   
   > 1625:0062 D3 E0          shl          ax,cl     ;*E0   
   > 1625:0064 30 25          xor          [di],ah   ;***25   
   > 1625:0066 47             inc          di   
   > 1625:0067 AA             stosb                  ;*AA   
   > 1625:0068 4F             dec          di   
   > ; test if cl has reached 8, if so reset to 0; "and cl,6" is too many FUs.   
   > 1625:0069 51             push         cx   
   > 1625:006A 58             pop          ax   
   > 1625:006B 34 30          xor          al,30   
   > 1625:006D 34 38          xor          al,38   
   > 1625:006F 75 03          jnz          0074      ;*03   
   > start_here:   
   > 1625:0071 4F             dec          di   
   > 1625:0072 50             push         ax   
   > 1625:0073 59             pop          cx   
   > 1625:0074 AC             lodsb                  ;*AC   
   > 1625:0075 36 36 71 C3    ss:jno       003A      ;*C3, corr:jmp=>03C   
   > ; (jmpback-amount reused as 'ret' trick!)   
      
   I see you scrambled your getnum. Well, now only 12 SMC points for B64   
   (11 for me) and 61 byte instead of the previous 16 points on your short   
   47 byte decoder this will become one byte less than my 209 byte variant   
   (+14-5*3 = -1).   
      
   >> Ok if just for the fun of doing it, but I still don't understand why   
   > []   
   > FSVO "fun"!   
      
   very obvious indeed :)   
   __   
   wolfgang   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)