(12-16-2009, 05:51 PM)StripTheSoul Wrote: [ -> ][color=#C71585]These games have been reported (multiple times) to have random crashes, so the crashes are not reproducible but it's obvious that Dolphin is having issues here:
- Baten Kaitos: Eternal Wings and the Lost Ocean (GC)
- Baten Kaitos Origins (GC)
- Mario Superstar Baseball (GC)
(freezes randomly during intermissions; even on single core)
- Monopoly Streets (Wii)
(when it crashes, it does so on loading screens)
- Mortal Kombat Deception (GC)
- Zelda OoT MQ and Collector's Edition OoT/MM
(both freeze intermittently without a direct way to reproduce)[/color]
So random crashes that can't be reproduced on demand = deterministic behaviour?
Who says they're random? Just because they seem random doesn't mean they are. I bet no one has tried to record a movie of it, and see if playing it back results in the same crash.
I tased spider man when it crashed constantly in JIT, and failed to even boot. And i did it in JIT. All i had to do was boot in interp, then switch to jit. Of course it still crashed randomly then, but it wasn't a problem. Because guess what? It wasn't really random. If it crashed, all i had to do was do something slightly different to avoid it, and then walah, it played back perfectly, magically avoid every "random" crash.
Tueidj: On netplay I can confirm those crashes are stable; both sides will crash in those games at the same point. Confirmed that in Mario Superstar Baseball.
I guess I don't know for sure about the rest, though.
(03-05-2014, 07:28 PM)JMC47 Wrote: [ -> ]Tueidj: On netplay I can confirm those crashes are stable; both sides will crash in those games at the same point. Confirmed that in Mario Superstar Baseball.
The premise that determinism only applied to netplay/movies was dropped:
(03-05-2014, 04:40 PM)RachelB Wrote: [ -> ]If you give the same input (which doesn't just include controller button presses), with the proper settings, the result will be exactly the same, every single time, with the only exception being if you use wiimote+save states, and possibly some specifically crafted homebrew program you might create specifically to prove dolphin is not deterministic. Prove me wrong.
It is stated multiple times that the crashes are not reproducible, so either they're non-deterministic or someone is telling fibs and the reporters need to be educated about how to properly test for reproduction.
(03-05-2014, 08:27 PM)tueidj Wrote: [ -> ] (03-05-2014, 07:28 PM)JMC47 Wrote: [ -> ]Tueidj: On netplay I can confirm those crashes are stable; both sides will crash in those games at the same point. Confirmed that in Mario Superstar Baseball.
The premise that determinism only applied to netplay/movies was dropped:
(03-05-2014, 04:40 PM)RachelB Wrote: [ -> ]If you give the same input (which doesn't just include controller button presses), with the proper settings, the result will be exactly the same, every single time, with the only exception being if you use wiimote+save states, and possibly some specifically crafted homebrew program you might create specifically to prove dolphin is not deterministic. Prove me wrong.
It is stated multiple times that the crashes are not reproducible, so either they're non-deterministic or someone is telling fibs and the reporters need to be educated about how to properly test for reproduction.
If you feed the emulator different inputs, you get different results. It's obviously impossible to feed the same inputs without using netplay/movies (as even something as small as start time could affect things, and both make sure that is the same). As per JMC, the reporters clearly need to be educated on how to test for reproducibility. Or i see your quoting a four year old post. Even if the issue was at some point non-deterministic, that doesn't mean it still is.
Correction: A post that was created four years ago and has been constantly edited to stay up-to-date, right up to yesterday.
Quote:Even if the issue was at some point non-deterministic, that doesn't mean it still is.
There's no way to know when those particular things were added, or last updated. More likely though, it was never actually tested. Whether because the tools to do so didn't exist at the time, or because i've never seen anyone else but me and jmc ever actually properly test if something that was seemingly random was actually random or not this way.
Gee I dunno, you could try actually reading the thread.
JMC47 was the one who reported "Zelda OoT MQ and Collector's Edition OoT/MM both freeze intermittently without a direct way to reproduce."
Ok, and in this thread:
Quote:I guess I don't know for sure about the rest, though.
Ie, he never tested.
I don't know why you would assume that because a bug reporter says there's no way to reproduce it that there's actually not. You put too much faith in people. You'll also have to forgive me for not wanting to search through a 29 page thread. But that just goes to what i'm saying: people are lazy. They see something doesns't work, and report it as such. They almost never bother to extensively test it.
tueidj: Yeah, I can't find a consistent way to reproduce that bug. But, for lulz, I decided to play that single player game on netplay and see if they both crashed at the same point. It took about 45 minutes for the game to crash, but, despite having the second computer about 2 seconds behind, they both crashed on the same frame.
I do not know how to reproduce said crash; but when it happens, it happens. You could probably learn more from shuffle2 or someone smart rather than me, who just pokes the emulator with a stick and hopes for the best.
Anyway, sorry for the confusion; just because there isn't an exact method to reproduce (I guess recording a Movie to the crash would work, except for the small little problem that you save a movie AFTERWARDS) doesn't mean it's inconsistent. I don't know what else to say, other than it's confusing. Netplay makes the emulators sync up very, very well, right down to crashes that I cannot trigger at will.