Bug submission thread (for RetroArch 0.9.9)

Can anyone confirm retroarch is working with psx ROMs on ios 7.1.1?

I’m desperately trying to fix a crash-on-psx-ROM-load problem. I’ve reinstalled retroarch, downloaded the scph1001.bin BIOS file(and changed to all lower case), placed it in var/mobile/documents/.retroarch folder. I’ve downloaded FFVII and Crash Bandicoot from coolROMs, placed in var/mobile/documents/ROMs, unzipped them to .bin files. Every time I choose a psx ROM it then asks me to choose core, I choose the psx ReARMed(only option as others don’t show up) but it crashes straight away. Also it might be worth noting that after it crashes, a pcsxr.log file is placed in var/mobile/documents, here’s what it reads;

Starting PCSX-ReARMed 0 Running PCSX Version 1.9 (Jan 11 2014). psxMap: warning: wanted to map @80000000, got 0x61f9000 Init new dynarec

Somebody save meh!

Cheers

I don’t know if this is a known issue, I scanned through the thread quickly but didn’t see it.

If you attempt to open an N64 game while rewinding is enabled in the settings, RetroArch will crash. This is in 1.0.0.2 64-bit on Windows 8. I’ve tested it by downloading a fresh copy of RetroArch from the site, running a game in it, then changing the rewinding setting and only that setting and attempting to re-open the game – boom, crash.

If you attempt to enable rewinding while a game is already open, nothing bad happens, but any attempts to rewind give an “End of rewind buffer” message, so clearly something’s going on. Maybe this isn’t even a bug but just a limitation of Mupen64plus, but since it’s something potentially frustrating for a lot of users I thought I’d mention it anyway.

@froggo Yeah, it’s a known issue. It’s definitely a potential stumbling block for people that aren’t already aware of it, though.

UPDATE: savestate serialization is fixed now, so rewind should work in mupen64plus in latest git. The serialized states are large, though, so be sure to have a large rewind buffer. Netplay should work, too, though the size thing may be a problem in bandwidth-limited situations.

Aside of the above (quoted as a reminder) is it possible next mame cores have the nag screen (that’s how it seems to be called) disabled? Also how is lightgun support in these cores?

I also want to take the chance to bring to attention the modifier bug although it has already a “ticket”.

Related my third quote above, here is an expanded explanation. Apart of Custom DAR, Integer/ratio Deviation could be a necessary evil for not integer scaling.

There is a core option to disable the nag screen and the warning screens already.

Thank you AndresSM, I suspected that might be possible but didn’t check myself, I read a recent complain about this and wanted to give it more voice.

I want to do a review of current issues I have so far spotted (and aren’t yet adressed), this can also be called the hall of shame since they aren’t likely to be fixed any time soon (lack of feedback): -Resolution Independent Custom DAR -Take full signal resolution for DAR calculations (and a crop setting for the extra padding if desired) -Mupen64 no directional controls when input_autoconfigure false -Autoinput axis takes precedence over manual button (of same physical input) when autodetect is false. -RGUI independent controls (no fun when you have to guess what is enter, back buttons, etc) -FBA core fails to load savestates. -NVRAM file being saved on rom path for FBA core, this was fixed a few months ago for MAME core -Not able to load redearth in fba core -Mednafen .mcr saves saving to system folder (from memcard1 onwards) when -s call is used (maybe the fact of… -…saving .mcr is something to be fixed, since we already have .srm libretro saves) explanation -Nestopia core renders black when custom ratio is loaded from cfg (index = 22) -RetroArch refuses to render image for PSX’s FFIX game (Spain region patched), it has audio but image is black, mednafen_psx core. -PAL version of N64’s Majora’s Mask doesn’t work. -famicom disk swap button is defaulted to “Shoulder L” and not “Y” using Nestopia. -yabause_libretro.dll the saturn emulator does not work out of the box (REDUMP images used). -There’s an issue where taking a screenshot while using the shader advanced-aa.cg the game is resumed, at least in mupen64.

I will add more with time.

“bugs” need to be confirmed to become bugs addressed. In both threads about your input issues we tried to help you and we tried to reproduce the “bugs” but we couldn’t you keep saying this is a developer issue when it’s not unless someone else can reproduce it.

input_autoconfigure doesn’t affect any particular core in any way. If the buttons and axes work in ONE core they work in all of them assuming the buttons are wired correctly to the retropad.

Please come into the IRC channel so we can try to solve this in realtime. It’s easier than this delayed discussion.

I have tried hard in order to be heard here, since nobody replied (only you about the psx issue) I decided to directly file it as a bug. The N64 joystick doesn’t work when autodetect false, because RA on a whim wants these (input_player1_l_x_plus_axis) to be configured instead of default (input_player1_right_axis). That is redundant data and may confuse many people, having two places for the same stick. About the PSX issue, it is indeed happening, and there’s a bug, I just haven’t had time these days to continue my research, I will forward you during the day. edit: I edited out the bug because i can’t reproduce it. Let me keep my findings on the other thread.

Digging on the button configuration mess RetroArch is, there is a hidden rule that axis take precedence over buttons when input_autodetect true.

For example by default C buttons for N64 are mapped to right joystick when input_autodetect false. If I want them mapped to d-pad I would write:

input_player1_r_x_plus_btn = “h0right” input_player1_r_x_plus_axis = “nul” (Auto: +2)

But it won’t work. As you might know manual configuration always take precedence over automatic input, so for me this is a bug. In order to make the above settings work you need to turn input_autodetect to false. Problem is, now you have to manual map all the buttons by hand.

edit: meh I better don’t

edit: much better.

Another bug, I updated my above post anyways. “input_enable_hotkey=b won’t work when RGUI, so for example I enter RGUI with b+F1, I can’t exit with b+F1”

I think controls in RA are a mess. I think default controls for RGUI should be d-pad (or/and joystick) and B and A buttons. And that the toggle key for entering RGUI (as well as the modifier that enable it) should work from the game and within RGUI.

btw, not necessarily a RA bug, but I can’t play Dino Crisis 1 on mednafen_psx it freezes on the advisory screen. (I use PAL version)

works here on a recent build

I think controls in RA are a mess. I think default controls for RGUI should be d-pad (or/and joystick) and B and A buttons. And that the toggle key for entering RGUI (as well as the modifier that enable it) should work from the game and within RGUI.

The toggle button works in RGUI and in-game, maybe you have a build affected by my last comment here: https://github.com/libretro/RetroArch/issues/748

Great!, can confirm, it must be fixed on these past days. I hope they can fix the controls within RGUI, in my opinion they should be fixed controls (not changing per core).

edit: edit because I found another bug, it was fixed for MAME but not for FBA. To save nvram files on save path and not in rom path.

I found a bug with game saves.

Games do not save immediately like they should.

I was playing Earthbound on Retroarch Android 1.0.0.2 on my Fire TV using Snes9x Next,and I made a lot of progress from StoneHenge to Tenda’s overcoming shyness. I then wanted to know where to get the Horn Of Life item so I could get Ness’ Hall of Famer Bat,so I Alt+Tabbed to search in a browser and found it. The game sound played in the background and my device was choppy because of the 4xBR shader at 4x resolution. Then I returned to Retroarch and resumed the game,only to find it invisible with a black screen while the game is still running. The menu was also invisible,which kept me from exiting normally,so I forced stopped it and opened it again and started Earthbound up. I had already saved it earlier in the Tenda Tribe room with Ness at Level 62,but when the save list came up,I was enraged to find him at Level 54 again,knowing that all of my progress was lost.

Please fix this MAJOR bug in the 1.1 update if it has not already been fixed. You only need to change how games save so they will immediately write the save file at the same moment the game normally would. In Earthbound’s case,when telling Dad to record your progress.

Because of that,I have to do everything again that I just did before,even though I told it to save the game several different times!

This is expected and intended behavior. There is an option in RGUI to automatically write SRAM to file (I think it’s called autosave interval or something like that) and it defaults to every 10 seconds or something.

The reason for this behavior is that some games use SRAM for scratch space and to write it to file every time it changes would not only crush performance due to I/O bottlenecks but it would also thrash your flash memory, as it would be using it as RAM. There has been some talk of introducing some heuristics to try and discern between normal saving and SRAM-as-scratch-RAM changes, but nothing has been implemented and it likely wouldn’t be foolproof anyway.

In the meantime, set the autosave interval and try to exit cleanly whenever possible.

The not saving issue can be worked-around by adding a manual save option in RGUI so you could make sure it saved. That way I could hit that option before attempting to switch to another app like a web browser whenever I need to. Or it can be fixed by adding a save flag whenever conditions change such as switching to a recent app.

The black screen issue is the actual bug that really needs to be fixed. As I mentioned already,it happens when switching to other apps using Alt+Tab (my “recent apps” hotkey) while the game is still running in the background,and returning to retroarch which gives me the black screen with the game still running.

iOS 7.1x, all retroarch versions.

PSCX-Rearmed is not working with A7 devices. Crashes on start load. The non-jb version is working, so I’d guess it’s a problem with dynarec on the 64-bit devices.