ObjectMerger bugs

[Xonok]

Over time I've found ObjectMerger has problems with several fields, mainly bitflags. It interprets them as strings or sometimes even something I have no idea what it is.

Example:
//! external ObjectMerger w3a ACvs A0Z1 Poi4 1 1 Poi4 2 "1"
This line treats both the 1 with and without " as strings. It would be logical for there to be some way of overriding that stupidity, since "1" is 49 as a string and 1 as a bitflag. I'd use the ASCII code for 1, except that it's not even printable!

Similar bugs exist with other fields too. For instance, death type likes to produce gibberish whatever you input there from objectmerger.

Is there any possibility of a patch?

 

[PurgeandFire]

The bitflag bug is mentioned here with a link to the hotfix:
http://www.hiveworkshop.com/forums/jass-ai-scripts-tutorials-280/lua-object-generation-191740/#Bugs

I don't remember the other ones you've mentioned. Try using Lua instead of the object merger and see if it has better explicitness.

 

[Xonok]

That one fixes channel only, not other bitflags. I have it.

I tried explicitly using objectmerger(as seen in the first post), no difference. Eventually I used a hex editor to hack the 01 (SOF) sign into the object generation and it worked seemingly right. Except that I don't yet know for sure, because I found the Envenomed Weapons ability also isn't fully levelable. The object editor value looks right(ctrl+d), but it turns out that Envenomed Weapons loses its stacking types if they are touched at all. Also, only the first level has them, so it's effectively unlevelable if you want it to stack. The damage works fine though.
I confirmed the things about Envenomed Ability by changing them manually, instead of with objectmerger.

 

[WaterKnight]

Did you try "\1" or string.char(1)?

 

[Xonok]

Did you try "\1" or string.char(1)?

12591 for the first one.

string.char actually seems to work right, surprisingly.

 

[WaterKnight]

http://www.lua.org/pil/2.4.html

needs 3 digits -> \001