Posts for Alyosha
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
I don't know how I did it, but somehow one of my pm's is marked as unread even though I read it, and now when I click on it I get this message:
Could not update private message read status for user

DEBUG MODE

SQL Error : 1690 BIGINT UNSIGNED value is out of range in '(`nesvideos_forum`.`users`.`user_unread_privmsg` - 1)'

UPDATE users SET user_unread_privmsg = user_unread_privmsg - 1 WHERE user_id = 8061

Line : 384
File : privmsg.php

Backtrace:
  includes/functions.php[743]: ErrorPrintBacktrace()
  privmsg.php[384]: message_die(202, {string(53)}, string(0) ""
, 384, {string(45)}, {string(91)})
I can't delete it either because it will give me an error message if I try to do that too:
Could not update user pm counters

DEBUG MODE

SQL Error : 1690 BIGINT UNSIGNED value is out of range in '(`nesvideos_forum`.`users`.`user_unread_privmsg` - 1)'

UPDATE users SET user_unread_privmsg = user_unread_privmsg - 1 WHERE user_id IN (8061)

Line : 960
File : privmsg.php

Backtrace:
  includes/functions.php[743]: ErrorPrintBacktrace()
  privmsg.php[960]: message_die(202, string(33) "Could not update user pm counters"
, string(0) ""
, 960, {string(45)}, {string(106)})
Any ideas how I can fix this? Other pm's work fine so it's not a big deal I guess, just kind of a nuisance.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Ah, yeah I should have thought of states when I was doing it as well. I added the new variables to the state now so let me know if it's fixed (remember that you'll have to delete the greenzone file in the .tasproj or else it will automatically crash from not matching state size.) Also I found a whole bunch of test roms here: https://github.com/AntonioND/gbc-hw-tests So looks like I have a lot of work to do! I'll look at the other commits after I sort everything out here and can look at exactly what is happening. EDIT: I finally found the bug in mode 3 timing during windowing and now I am syncing up with Gambatte. Problem solved! The new test roms also revealed a couple other minor bugs so things have been really productive all around. Incidently, the test halt_if relies on the same halt bug that is needed in pokemon yellow, so this information was out there (although obscured) 3 years ago! Kind of funny how information is just strewn about out there.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Yes I just converted the VBM. Then I had to adjust the start point a bit for loading (and BIOS of course.) After that the gameplay was identical except for the very end where I adjusted the input by random guess and check until I got the glitch. I briefly tested SML1 and it does look like it would sync with only adjusting for load times (I made it through the first few levels anyway.) Well, the movie as it is published now is definitely not accuracte, since the current inputs do not even come close to hitting correct timing, so in that sense yes VBA is bugged, but the new movie still needs to be confirmed on console somehow to be sure.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Uh oh , I was really hoping it would just get to the end without issue. Yeah please do try Gambatte and see what happens, I’ll look at it this week and see if I can anything funny going on. EDIT: I manually resynced the gambatte version to see if things diverge anywhere, it turns out they diverge at 177475 (GBHawk) or 178390 (gambatte). This is right when you are using cut to clear a path after doing a bunch of shopping. Once again the point of divergence is mode 3 timing. I seriously need a test rom to figure out what's going on there. Of course I don't know for sure if what gambate is doing is correct either, I'll need to wait and see what TiKevin83 comes up with.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Nice work re-doing this so quickly. That second boss fight looks so smooth now! Yes vote!
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Ah, thank you that script did help in finding proper frames. With it I was able to get a game end glitch: http://tasvideos.org/userfiles/info/47098069959194767 It works in both gambatte and GBHawk, so it's pretty likely it also works on console. Unfortunately, this is only in DMG mode. I haven't found even a close hit in GBC in GBA mode (the one where console verification is easy to check.) Maybe it's possible to directly check on a regular GB but I don't know, seems pretty likely though.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
MUGG wrote:
I verified the pause glitch also works on Bizhawk (gambatte core). $A201 onward are executed on rare occasion when doing the bug, and depending on the values of those addresses, various things can happen, as was the case in VBA. I didn't manage to do the credit warp by editing memory to C4 C3 05 ... C0 though. I only did manage to leave the level early by C4 BC 05 ... C0. This could be because of the glitch being stubborn or because of emulation differences, I don't know. I tried with five different scenarios and it barely works. I remember the bug was more compliant in VBA.
MUGG, do you still have a bk2 of this? I'm interested in seeing if this can be made to work now that there is a fairly robust console verification pipeline available.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
TiKevin83 wrote:
I should also mention I'm getting a lot of audio problems with GBHawk right now, it seems to speed up and slow down a lot which in turn makes the pitch sway wildly.
That's because during loading when the screen is disabled multiple audio frames are squeezed into one input frame. When made into a video encode and played back it should sound fine since multiple video frames will be used as needed to account for the length of the audio sample.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
I'll fix the screen color (I just set it to black when I first wrote it and didn't update it at all.) It's slower then Gambatte primarily since it's written in C#. It's probably as fast as it's going to be. My goals are accuracy and maintainability, not speed, although I do try to optimizae where I can. I realize this is a pain especially for longer TASes. Honestly as long as you know it works on console you might as well stick with Gambatte, especially if gifvex continues to be able to keep it updated. Maybe somewhere down the road GBHawk will really start to pull ahead in terms of accuracy, but I don't think we're there yet. Still cool to have Pokemon Yellow syncing though, let's go for a full run!
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
I just committed some improvements to GBHawk that allow the game Airaki to work. This is a game by Furrtek, which is known for their anti-emulator checks in their games. This game did 2 things. First, it checked that reading from the Cart was disabled while DMA is in progress and DMA address is from ROM space. Second, it checked if the timer divider register is being used to clock the length in SQ1 in the audio. (You can prevent clocks by constantly resetting the DIV register.) These are both well known behaviours, but aren't commonly implemented in emulators I guess, at least at the time the game was made, Sameboy for exmple runs it just fine.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Gambatte has gotten a pretty important improvement today with a newly discovered glitch in Halt behaviour being accurately emulated. This glitch is relatively uncommon but it's impact can be fairly large, since it basically causes a second Halt to occur. So if anyone out there is working on a GB or GBC run in Gambatte, I would strongly encourage you to check out the dev build and see if it effects your run. The recent Pokemon Yellow run desynced on console due to this glitch in an otherwise unremarkable section of normal bike riding, so it is an important accuracy fix to be sure.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
feos wrote:
TASTUDIO USERS PLEASE TEST AGAIN I tried to fix stuttering leftovers, please give it thorough testing once again! Also fixed accidental removal of state for last edited frame. Link to interim build is in the OP.
I was having some trouble with stuttering that seems to be fixed now, it occured when going to back to the start of a movie when the current frame was several tens of thousands of frames later. I'm guessing these commits fixed it so thank you feos.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
http://tasvideos.org/userfiles/info/47030679356753062 Here is 3-10 in 422 steps (4 better then Nitrodon's improvement to 426.) That's probably about all the effort I'm going to put into this game. If GBHawk becomes acceptable in the next BizHawk release I'll likely just submit what I have.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
I've been struggling through trying to implement new information found by gifvex. So far I can pass a new test ROM that is supposed to verify this behaviour, but ultimately I don't match the expected console behaviour in Pokemon yellow for some reason. I did re-organize the CPU to poll the INT line on the correct cycle though, so that's good since it needed to be done eventually anyway. Unfortunately though this didn't actually solve any problems, oh well. So if anyone notices any regressions in the latest dev builds do let me know.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Nice run! It's god to see another run brought up to speed with RTA. Yes vote.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
http://tasvideos.org/userfiles/info/46919445429283957 1 move saved in 3-9. (I also improved the intro menuing.) I looked over the other levels but couldn't find anything, so it's on to 3-10. If anyone else decides to try to run a solver, I would be quite interested in the results for some of the higher move count levels, although some of them do seem to have only one 'fast' solution path.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
4 to go. Glad to be done with psychic world that was a grind.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Mothrayas wrote:
I'm not getting the movie file to sync, using BizHawk 2.2.2 and the same ROM MD5 as in the movie header. The movie file has no sync settings. Is there anything in there I may be missing?
SMS -> Display Type -> NTSC I supposed to fix that for 2.2.2 but never got around to it.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
mklip2001 wrote:
Any chance at an encode? I really like this game, but I haven't been able to get emulators set up. No hurry though.
Sorry but my net right now is too slow to upload this long of an encode. Although, it's visually almost identical to the original, so there isn't much point really.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
http://tasvideos.org/userfiles/info/46805336826281923 I found improvements in 2-1 and 2-4. I'm thinking the improvements cybershadow found were these plus 2-6, so that list of 5 levels would be: 2-1, 2-3, 2-4, 2-6, 3-2 There is also the improvement I found in 2-7, and the knwon improvement in 2-2. So that's 7 levels improved so far. I looked over most of the other low move count levels but didn't see any obvious improvements. So the only other known improvement is in 3-10 which I haven't looked at yet. I'll probably look over the other levels up to that point first.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Config -> Speed/Skip -> (set the speed you want)
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Huh, not sure what to suggest then, maybe you have a bad rom? That message pops up when a rom is in Quicknes’ blacklist, but I don’t see Mario 3 there
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
thommy3 wrote:
CyberShadow had a bot search for better solutions and mentioned it in this thread. See my old post for a 2-2 tip. It might be exactly that position the screenshot is showing. Then going for the left 3x1 block will be even faster.
Ha, nice catch, I didn’t even see that! So someone already coded a solver for this ... 9 years ago 0_0. Man, there is so much lost or abandoned work out there, it’s kind of sad. EDIT: I managed to figure out the improvement in 2-3 to get 98 steps. I implemented the 205 step route in 2-2. I also implemented the 2 step improvement in 2-6 from thommy3. I haven't looked at the 3-2 yet which cybershadow listed as 180 steps (5 step improvement.) Nitrodon also mentions an improvement of 3-10 to 426 steps that I haven't looked at. CyberShadow originally mentioned 5 improved levels I don't know if the 2-6 and 2-7 improvements are part of those, so at best 1 more improvement is out there and at worst 3. EDIT2: http://tasvideos.org/userfiles/info/46783748064223941 It took some effort, but I found 180 moves in 3-2. Only one more (stated) improvement to find.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
Here is a movie file that syncs in GBHawk in the current Dev Build. I included the above time save, and also found a 3 step faster route in 2-7. I might spend a little more time looking at the other levels, but so far nothing else has stuck out at me as being obviously improveable. Also Kwirk does some stuff with audio mixing (cutting off the wave channel mid sample) that results in some obvious popping. I'm guessing there is some kind of filtering I should be doing to mitigate this, but I haven't looked into it yet.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3602)
Joined: 11/30/2014
Posts: 2752
Location: US
arkiandruski wrote:
Good news. Switching the core to NEShawk didn't desync the movie at all. Continuing with new core. But if NEShawk is more accurate, why is QuickNES the default?
Oh nice, glad it synced up for you, that's lucky. QuickNES is like 20x faster so is more accessbile for regular (non-TAS) users and those with weaker hardware.