I would guess that DOSBox's DOS implementation is to blame, crashing instead of gracefully handling an invalid filename.
custom-1998-10-04-4map.zip contains a copy of CUSTOM.EXE which is 100kB too short, it doesn't run. GAME.EXE and other files seem to be fine.
custom-1998-07-27.zip has a working CUSTOM.EXE, though sometimes it crashes in DOSBox. It seems to depend on the .rpg file used. E.g. accessing an enemy's General Data in NPC_TAG.RPG crashed, and sometimes adding a new map does as well.
I hear that CUSTOM.EXE from custom-1999-12-21.zip is less crash prone, but I'm going to try using 1998-07-27 instead.
Feenicks wrote that that (not-4-map) 1999 version is OK to use for that reason, but I presume you're still meant to stick to 1998 features?
Here's a few things I noticed from that version, figured I'd post them here since they're relevant to the current contest...
If you look at whatsnew.txt in a nightly build you'll see I added 1998-10-04 and 1998-07-26/27 to the bottom of the file for easy searching. In particular:
* Textbox hero swapping conditionals were added in 1998-10-04
* Caterpillar parties added 1998-10-04
* "NPC positions are preserved when you fight a battle" was 1998-10-04... yikes!
* Font editing was added June 18 1999
* Vehicles added May 8 2000
* Item and hero auto-set tags added June 18 1999
* Harmtiles added January 25 2000
* "Yay! The mapeditor in wallmapping mode no longer flickers in a seisure-inducing way. Yay!" -- January 25 2000
* Poison, stun, regen were added Nov 23 2002. I think I actually remember that.
Also I wrongly said that there are no enemy bitsets in 1998-07-27. Turns out I needed to scroll down past a bunch of blank bitsets!
Nathan Karr wrote:
I tested the elements extensively in 2003/2004 with characters who did flat 100 damage and Do Not Randomize bitset. Weak To was 200%, Strong To was 18%, Weak To + Strong To was 36%, and of course Absorbs would invert the damage into healing.
Fnrrf was right, "strong to" was 12% damage, "weak to" was 200%, and both was 24%. "Bonus vs [enemytype]" was 180% damage.
Nathan Karr wrote:
I remember learning later that type-slayer weaknesses were x1.8 instead of x2. I didn't experiment with those much because I didn't find it very useful to have "fail if enemy is [type]" compared to nowadays where I can have "fail if enemy isn't weak to [type slayer] minimum 101%"
You've warmed my heart. Ever since I replaced "fail if enemy is [type]" with the percentage thresholds I've regretted it, thinking noone actually wanted that change.
I would do things differently today. I think I would keep around separate old obsolete settings separate rather than trying to find some way to convert them to an over-general new system.