Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Ubercapitalist wrote:
My verification run is complete. Thankfully I only needed to hex in a couple button presses do to additional text lines at the start, but the rest syncs perfectly.
The verification run consists of three DTM files due to my inability to sync over disc swap and resets.
1. Main Game Disc 1
2. Main Game Disc 2
3. Unlock Wesker
Should I create one new submission containing all three files (my preference), or a separate submission for each DTM?
Upload those three to userfiles and link from this submission.
And if one needs to hex this submission for it to sync with the verification movies, upload that too and request replacment.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
adelikat wrote:
Just pointing out that this movie exploits unitialized ram
AFAICT, it does not. It neither enables random initialization nor has initial RAM contents.
And indeed, loading the movie and hashing the WRAM gives the usual result.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
synnchan wrote:
On games like Metal Slug, if you press F2 you go to a menu that lets you change a lot of things. I want to change the game's level from 4 to 8 and enable blood. The game changes normally if I do that, but if I start recording a movie and put "record from power-on", the game returns to level 4 and without blood.
How can I record from power-on while keeping the changes made through F2?
At least one Metal Slug movie enters the setup menu somehow from the movie and adjusts settings there.
So looks like there is a way to rerecord the effect of F2 key.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
I once looked at internals of PCSX2. It seemed that due to how the emulator core <-> gamepad plugin interface worked, it seemed like one would have to do input recording at gamepad plugin.
But the PCSX2 core probably does not send the needed events (in practicular, savestate/loadstate with option of saving/loading extra data) to the pad plugin.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
feos wrote:
Manipulating memory and code is basically the very thing one needs to do to glitch a game to the ending. But it doesn't mean one has total control on the game's execution. Only the run that obsoleted 1945M used total control explicitly. Otherwise, most game end glitch runs we have just confuse the execution pointer to jump to the ending routine, not command it entirely.
AFAIK, 1945M poked a routine that triggered the credits (and it was substantially more complex than just jumping to one address) into OAM mirror and then caused the game to jump to it.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
jlun2 wrote:
Well, there's the previous movie. Not sure if it can lead to total control, but it corrupts memory.
Even that method involves CPU jumping into WRAM and executing stuff that it should not.
AFAIK, there's also third way to perform credits glitch. I have no idea if it invoves CPU jumping into RAM.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Ilari wrote:
Don't offhand remember how I stubled upon the SA1 emulation bug.
Oh yeah, I saw that different builds executed seemingly different instructions and then did go to look how the heck the actual instruction it was executing was determined.
Turns out, it was determined by crap left in one CPU register at one point in code.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Bobo the King wrote:
This problem has been on my mind a little lately. I haven't given it a whole lot of thought, but to the extent that I have, I'm stumped.
1) Construct a function of two real numbers that, using only addition/subtraction, multiplication/division, and absolute values that takes the value of the larger of the two numbers. (So construct the max function using only absolute values. This should be very easy.)
2) Construct a function of three numbers with the same requirements as above (using only a/s, m/d, and absolute values) that takes the value of the middle number. (This is only a bit harder.)
3) Construct a function of four numbers with the same requirements as above that takes the value of the second-largest number. (I don't know off the top of my head how one might do it.)
Edit: I found one way to do it. It's a little tricky and I'm sure there are many equivalent ways. I also suspect the process can be generalized to yet more variables.
One fairly brute force way to do it is construct sorting network and then implement it using min()/max() and further using allowed primitives.
But without Common Subexpression Elimination, the complexity of resulting expression likely grows exponentially in number of inputs.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
SmashManiac wrote:
I'm happy to see that this was caught, but how were you able to figure out that there was an emulator bug in the first place? Was it because you tried syncing the movie on a different build and failed, or was there a different reason?
Got a report that there was trouble syncing it on the latest beta build and went to investigate.
Don't offhand remember how I stubled upon the SA1 emulation bug.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
jlun2 wrote:
Fortranm wrote:
It is the game's fault. I tried Bsnes v84 and Higan v95. There are sound glitches even under Snes9x 1.52 & 1.53. Snes9x 1.51 is my final decision since the game doesn't seem to give correct sound under any emulators with higher accuracy.
Wouldn't a run that does that be rejected for (known) inaccurate emulation?
No. Because it doesn't materially change if the run is possible on real hardware (it is just sound glitching).
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Patashu wrote:
I think it's not so much 'it depends on emulator issues', so much as 'The setup isn't the same as it would be on console, in a way that differs wildly between different versions of the emulator, so the emulator needs to get more accurate first'
Basically, it is a bug in bsnes core (the handling of SA1 open bus reads is definitely incorrect[*]).
It is so buggy that even taking the exact same sources and compiling with different compiler version can break sync.
It is likely that the build (not just version) Masterjun used is the only one where it syncs.
[*] I have two different candidates for the correct behaviour:
- Do it like bsnes v094 does: Return MDR, updated on write.
- Do it like bsnes does for S-CPU (S-CPU and SA1 share execution core): Return MDR, update on read and write.
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
feos wrote:
BTW, what about this movie's branch? Seems it's some huge glitch finishing the game several minutes faster than expected. From the submission text, looks like a "game end glitch" to me.
What is the "proper" way to end this game? Is it reaching the top (after lots more wandering) or something else entierely (and the top also triggers credits for some odd reason)?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
TASPlasma wrote:
I was actually planning on doing a very short test run with a real Wii Mote/Motion plus, then manually adding in any frame perfect inputs.
That seems rather nasty to do, at least without writing some helper programs to decode/encode DTM files or somesuch.
Traditional hex editing does not work well with DTM files (because of the variable stride).
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
Tub wrote:
Can you elaborate what you're trying to do? There's a different set of requirements for, say, user tracking vs. user authentication, for whether you control the first party site or not, and how frequent page reloads are in the first and third party pages.
AFAIK, this is for user authentication, and there is little control of first party site (it frames/embeds the third party site).
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
WST wrote:
I might try; though, I can’t give any ETA yet. Probably in a few days (it takes incredibly much time and disk space to process even a single level/act, despite the automation, and also it’s very easy to make some mistake while dumping frames).
What takes so much disk space? The raw dumps? The intermediate files? The final encodes? Any reason why intermediates can't be done using FIFOs?
Emulator Coder, Experienced Forum User, Published Author, Skilled player
(1142)
Joined: 5/1/2010
Posts: 1217
martandelus wrote:
i'm not too upto snuff with saturn emu, so any feedback woiuld be boss, i think the one i tried was something called Yabuse.
If you can't get the game even running properly in newest Yabuse, forget about rerecording it.
I think there are better Saturn emulators than that, but none are rerecording-capable, have forks that are rerecording-capable nor have cores imported to any rerecording-capable emulator.
