Dynamic JASS hardcoding: Need Testers.

[Tom Jones]

OS: Windows XP Service pack 3.
Warcraft: v1.23

Test with plain editor:
Fatal, referencing bad memory.

Test with NewGen v1.5b:
Fatal, referencing bad memory:

Test with NewGen v1.5b with JassHelper v0.9.G.1:
Same story.

 

[Jesus4Lyf]

So this really does randomly work for most, but not for some?

>After I saved the map with JassHelper 0.9.G.1, the numbers stopped working but multiple phoenixes started spawning at the position of the original. It eventually crashed. I attached the new map to this post.

That's because it compiles differently, I think it spawns one or two less functions. You may notice that in my "addCall" function, it passes in a hex code and a parameter. The hex code is the internal function identifier of WC3 which I have to find by a memory search. I was trying to figure out a way to do this dynamically like I did with the array. But I can't think of any except writing a precompiler.

So basically, on the original map, if you change one of the function calls to 0xEDF (off memory) you get the same effect, phoenix spam. I'm pretty sure this is simply running the init function or something.

>As for making it RoC, I think removing the Phoenix would do it. Hopefully.
For the same reason I mentioned above, this would not work. The blizzard.j files are different lengths, so my relative bytecode pointer would almost certainly point to the wrong place, causing a fatal error.

See, from what Tom Jones has said, I can't see any reason why this should randomly fail on his computer...

And wow, azlier even got to save it! We must have the exact same development environment.

 

[kingkingyyk3]

My Wc3 crashed.

Windows XP SP2
Warcraft 3 TFT 1.23

 

[Tom Jones]

The reason for failing may be background applications?

 

[emootootoo]

worked for me unsaved and then worked for me saved aswell

had the full output both times

latest patch
win xp pro
saved with JassHelper 0.9.E.1

 

[Azlier]

>We must have the exact same development environment.
Evidently. I'm worried, now.

You're insane, you know that? This is what madness you come up with? You really are trying to smash Warcraft to tiny bits, aren't you?

 

[Jesus4Lyf]

>worked for me unsaved and then worked for me saved aswell
Congratulations. Thanks for the test. Seems if people save it on 0.9.E.x that it works. Makes sense to me.

>The reason for failing may be background applications?
Wish I knew more about how memory works. My best explanation would be computers not having enough RAM spare so it may write some things out to the harddisk, changing the distance between the anchor and array memory location. But honestly, I only learned what I needed for making this about 2 days ago, and until them I've never even looked at bytecode/internal command representation. XD

>You're insane, you know that? This is what madness you come up with? You really are trying to smash Warcraft to tiny bits, aren't you?
I was waiting for that.
EpicYes. :thup:

Spoiler

On the other hand, I found a legitimate use for the return bug!

At least when people start arguing over timer systems, I can say I created one that's more efficient than a hardcoded loop. Actually, it's probably much faster (percentage wise). o.o

 

[Azlier]

>0.9.E.x
...

I'm using the latest Jasshelper...

 

[saw792]

Sorry I took so long to reply.

I use Windows XP SP3 with Warcraft III:TFT v1.23, obviously JAPI and Grimoire disabled.

It displayed all the function call text, and also worked after saving.

I'm also using the latest JassHelper. Changes to inlining functions in 0.9.E.x onwards may be the reason why it works...

 

[Jesus4Lyf]

Hm. Odd.

The important part isn't that it displays the "Adding call ...xx" stuff. The important part is that each second it spits out:
SomeFunc
1
SomeFunc
1337
SomeOtherFunc
577

You got that, right?

 

[kingkingyyk3]

Actually, Why are you creating risky system??? I prefer you use other system instead of using the "Crashing" system. Although it is faster, but who wan to take risk to use that system??? I prefer stable more then risky.

 

[saw792]

Hm. Odd.

The important part isn't that it displays the "Adding call ...xx" stuff. The important part is that each second it spits out:
SomeFunc
1
SomeFunc
1337
SomeOtherFunc
577

You got that, right?

Yes, it displayed all the SomeFunc data every second.

 

[Jesus4Lyf]

>Yes, it displayed all the SomeFunc data every second.
Thanks.

Actually, Why are you creating risky system??? I prefer you use other system instead of using the "Crashing" system. Although it is faster, but who wan to take risk to use that system??? I prefer stable more then risky.

You may notice this is under the Jass Help section asking for testing, not under the Systems section saying "Use this awesome system".

I don't care about the system. People aren't just testing some system here, they're testing a princible: Is it stable to execute bytecode out of arrays in Warcraft III?

The answer? "Maybe." :thdown:

Hell, I'd use Key Timers 2 over this any day, but if this was stable then I intended to develop it a bit further.

Spoiler

Here's the real deal. People may suggest that module timer loops like Periodic Module are faster than KT2 and therefore KT2 should not be used. I say the interface is poor and the speed advantage isn't significant. Now I can say "Well, if interfaces aren't important to you and you want speed, go use a bytecode timer system which is even faster than your loop".

I don't really care enough to mean it, but I had to try it to see if it works.
Just for kicks. :thup:

 

[Viikuna]

Hes right. Testing crazy stuff is something without we would not have any of this cool stuff like return bug and H2I.


I must say I respect all these crazy researches and thingies you post in these forums.

Dont stop, keep doing it. :thup:

 

[Vexorian]

Since it does function calls, it is still not the fastest timer system on earth. Probably tied with PeriodicModule a manual loop is always better though.

I think there are more interesting and useful things to do with a byte code injector than making a fast timer system, however I can't picture them without some tool aid.

This trick to run bytecode is actually amazing, though makes me think of blizz as naive programmers, not feeling safe if there are other similar hacks possible that make you run machine code... It is strange/lame that it is not working for some people though.

So, there's theorically a way to have 8191*4 bytes of code makes you think of 'stuff' , if only there was a fast way to run a code variable...
I wonder if 20640 bytes of code are enough for everyone... Will have to ask pipe.

 

[Akolyt0r]

win xp3, warcraft 1.23

didnt even open the map with editor .. still instant crash on somefunc(1)

i highly doubt some kind of system which works using "bytecode" will work for Mac / Linux users ..

 

[Pyrogasm]

For the hell of it I tried. Worked neither before nor after saving.

OS X, Frozen Throne version 1.23

 

[Azlier]

Huh. Always worked for me. Win XP, SP2.

Over at WC3C, somebody apparently used bytecode outside of Jass to disable array bounds checking. This is relevant to my interests.

 

[Jesus4Lyf]

>I wonder if 20640 bytes of code are enough for everyone... Will have to ask pipe.

Not by my calculations. Two arrays, however, should be able to contain maps of up to around 30,000 lines.

So tell me. Could this be used to inject natives at runtime?
(Aside from the obvious that it could be used to wipe your computer.)

So like. EPIC BLIZZARD FAIL! :banghead:

Ahhhahahahahahahahaahahahahaha........

No qualms with the return bug being fixed here!
Hey imagine, DotA would even need to run its spells on a good system like KT2, and it would stop lagging! :O

This is what madness you come up with? You really are trying to smash Warcraft to tiny bits, aren't you?

PS. So how bad could Blizzard get sued for this if it was exploited (or could they)? =X

 

[Viikuna]

DotA doesnt lagg for me, but eh, what did just happen?

So it is unsafe to play public games now?

Now, I shall use this smiley to express my confusion:

edit. Shee-it. Tested toadcops test map.