Even with -dialect sjasm I still have errors.
I think your parser has some trouble handling spaces/tabs and their combination.
line: MAIN call GAME errors: ERROR: parseRestofTheLine: Cannot parse line main.asm#36: MAIN call GAME
I do not have any project on line.
Ah!! interesting line indeed!!
It's not the spaces/tabs. That is handled fine. The problem is that MDL's parser is thinking that the line is a macro call with a macro called "MAIN" and a parameter "call". Because the label does not have a colon. Since sjasm parses this line fine, I am making a note to allow this line when dialect is sjasm! We were just talking about this yesterday in another thread. Labels without colons lead to very ambiguous syntax and each assembler disambiguates in its own particular way. But this is very useful! Since my work is to figure out exactly the way in which each assembler disambiguates each of these cases, to mimic it in MDL 
So, finding these cases where I got it wrong is great!
I would be curious to see the line that causes that exception btw if you don't mind sharing!
MDL crashed there, so, indeed there is no error line. But there are hints (that probably only I can recognize haha) in that earlier error where MDL believes "speed_max" is undefined, and that the error happened in the "instantiateMacro" function of the SjasmDialect class, which means that was during the expansion of a "repeat" macro (which is the only sjasm-specific macro I currently have implemented). So, probably if there is any "repeat" that involves "speed_max", that'll be where MDL had troubles
In any case, I have just added some check around the line where the exception was thrown to capture the case where that null pointer exception can happen and give a meaningful error with an assembler line number. This together with many other improvements I have been working on will be added to the next alpha version!
Again, thanks a lot for testing, and please bear with me until MDL becomes robust enough, it's been only about 5-6 weeks since I started coding it, so it's still definitively rough
I see, interesting, thanks a lot! I'll check if there is anything funny in the parser with any of these cases. I suspect that the problem comes from the fact that when selecting "dialect -sjasm", I turn the parser into "eager mode" instead of the "lazy mode" (the default). In eager mode, the parser expands the macros right away, as they are being parsed, and thus, all constants involved (like speed_max) need to be defined by the time the macro is parsed. Maybe compute.asm is parsed before game.asm, and thus it is not yet defined. This should not be a problem though. So, maybe I need to make my "eager mode" a bit less eager
I'll investigate, thanks a lot again for sharing! (the other problem you mentioned above is already fixed, and should work well in the next release of MDL!)
@Metalion: with the lines you shared, I was able to create a test that reproduces the crash you report! So, I can now debug it Should be fixed later today!
@mzoran: of course! Something that would help me a lot is if there is any available winape-format project available online somewhere, so, I can test with it, and I could add it as another supported dialect.