[garbear]

Hmm, so there is no way currently to keep logging until the "real" error occurs?

No, the query ID things are all callbacks from the emulator code, can't add logging to the compiled code. i'll give it some thought.

[garbear]

@garbear you need to remove the trailing slash from systemdirectory variables (at least with genplus). Its making the check fail, I had to edit the addon setting manually to get segacd games to work.

http://pastebin.com/RS0Vzf8q

this has been fixed in the PR branch. i'll compile binaries/images when I have time (might be a few days)

[topfs2]

The query ID=15 isn't an error, just a component that I haven't implemented yet. My non-implementation (i think i just return NULL or something) is consistent with the libretro API, and the emulators must have forgotten to check for NULL, expecting the component to be fully implemented. This is most likely, as I have already patched one emulator for not checking for NULL properly.

However, last time a query ID error popped up, it was a non-error like above, and the game crashed in the game loop (but the query ID error was the last line to be logged). I'll resolve this by debug-logging right before the game loop is first run.

Bit OT, I configured a kill xbmc button on my remote (killall xbmc.bin) which used to work great with default openelec, however it restarts the whole machine when used with your build (which makes it kinda pointless,,,) Any Idea how to fix that?

My guess is that OE 3.2.0 changed, so that if xbmc crashes the entire system will restart. Seems like a valid strategy in any case. A better way to accomplish this would be to send a quit event to XBMC (you have multiple options)

I'm pretty sure I've found why this is happening, I placed a comment in your PR about it.

It seems like you are setting value = NULL when it has given you a value. And since you return true in the end of the callback, the emulator will use that value (no matter if its null or not)

[n1md4]

I installed OpenELEC 3.2.0 General x86_64 (is that the right version for an AMD CPU and nVidia GPU w/VDPAU?), select a SNES ROM (using bSNES Accuracy) and it all seems to work well. I had one question, I've an Onza Razor USB controller, is there a way to get this working automagically with emulators; as yet, without trying very hard, it doesn't work at all, shy of initiating --which is a good start I suppose

I'll continue testing ROMs and will post debug logs when necessary.

[ilm1]

OK, so here we go for a little feedback

I am unable to use ROM collection browsers because I can't find binaries of the Emulators to link them with RCB.

My PS3 Bluetooth controller is not working at all, I used to pair it using a tool but can't find it anymore. Anyone using a Bluetooth PS3 controller like me ?

In linux you can either use the sixad driver or the ps3d.py script included with xbmc.

It took quite some work, but the following configs work quite well (for an azerty keyboard!):
My sixad config: http://pastebin.com/u3kmWkNw (ps3 controller configured as input device)
Then for xbmc retroplayer I make sure the same keys are mapped:
http://pastebin.com/acPwpmsK

This works really nice (cave story, mame, couldn't try sega stuff yet due to bugs, possible related to the 32bit emulators I'm using).

Ideally I would like to use my ps3 controller as a joystick, but xbmc doesn't yet support on-the-fly joystick detection (meaning the controller must be connected at startup, and if it ever disconnects I need to restart xbmc). I also tried ps3d.py instead of sixas but then my controller starts rumbling and never stops (bug ?) Also then it seems I need to configure a keymap.py (instead of the traditional .xml files system).

[ilm1]

I'm pretty sure I've found why this is happening, I placed a comment in your PR about it.

It seems like you are setting value = NULL when it has given you a value. And since you return true in the end of the callback, the emulator will use that value (no matter if its null or not)

for what it is worth, it crashes inside check_variables of the genplus core:

Code:

[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/libthread_db.so.1".
Core was generated by `/usr/lib/xbmc/xbmc.bin --standalone'.
Program terminated with signal 11, Segmentation fault.
#0  0x9e64298f in check_variables () from /home/xbmc/.xbmc/addons/gameclient.genplus/libretro-genplus.so

Thread 15 (Thread 0xb2694b40 (LWP 19934)):
#0  0xb6b50b54 in pthread_cond_timedwait@@GLIBC_2.3.2 () from /lib/libpthread.so.0
#1  0x085014aa in bool XbmcThreads::TightConditionVariable<bool volatile&>::wait<CCriticalSection>(CCriticalSection&, unsigned l$
#2  0x085bb44c in ActiveAE::CActiveAE::Process() ()
#3  0x08da71be in CThread::Action() ()
#4  0x08da747f in CThread::staticThread(void*) ()
#5  0xb6921101 in ?? () from /usr/X11R6/lib/libGL.so.1
############# END STACK TRACE ###############

I can compile the genplus with debug information and check on it, if that would be useful.

In addition, the snes9x-next core fails to load. In the log file I find:

Code:

20:34:49 T:3008923584   DEBUG: SECTION:LoadDLL(/home/xbmc/.xbmc/addons/gameclient.snes9x-next/libretro-snes9x-next.so)
20:34:49 T:3008923584   DEBUG: Loading: /home/xbmc/.xbmc/addons/gameclient.snes9x-next/libretro-snes9x-next.so
20:34:49 T:3008923584    INFO: GameClient: Loaded SNES9x Next core at version v1.52.4
20:34:49 T:3008923584   ERROR: GameClient: <extensions> tag in addon.xml doesn't match the set from DLL value (smc|fig|sfc|gd3|gd7|dx2|bsx|swc)
20:34:49 T:3008923584   DEBUG: SECTION:UnloadDll(/home/xbmc/.xbmc/addons/gameclient.snes9x-next/libretro-snes9x-next.so)

The addon.xml also specifies zip, lets see if removing it fixes it... Maybe scummvm has the same problem.

[garbear]

I merged topfs2's PR (https://github.com/garbear/xbmc/pull/12 ) which should fix the settings crash for now.

[ilm1]

Thanks, genplus works somewhat now.

How are normal savegames supposed to work (for snes9x-next in particular) ? I can save in chrono trigger but whenever I quit the savegame seems to be gone (or I just don't know how to load it...). I suppose snes9x-next should create a memory card of some kind somewhere (.srm file) right ? However I can not find anything (tried setting a system directory too), is this yet to be supported ?

[garbear]

Yes, this will be supported in the future. I have a development branch somewhere that integrates a full-blown savestate manager. Memory card saving hasn't been implemented.

[ilm1]

K thanks, I knew about the savestates, but couldnt find any information on normal save games. I guess will compile the full branch to start my playthrough instead of the PR.

[n1md4]

I appreciate the development nature of the provided test builds, but i have a question, is it possible to update these builds to include the latest coded feature / fixes, or is a reinstall the only way to stay updated?

[Dreded]

First off thank you so much this is the closest iv ever been to getting emulators to work nicely in OpenELEC

Was wondering how to map keys in retroplayer I can see in the log that it receives them but does nothing(for the most part) for example if I press start on Xbox360Controller the log reads:

13:40:52 T:140130299217728 DEBUG: Joystick 3 button 8 pressed
13:40:52 T:140130299217728 DEBUG: -> RetroPlayerInput: Controller=3, button down=8, Action JoypadStart, id=3
13:40:52 T:140129851127552 INFO: LibretroEnvironment query ID=17: GET_VARIABLE_UPDATE
13:40:52 T:140130299217728 INFO: Previous line repeats 8 times.

but in both SNES-Next and FCEUMM nothing happens the game continues on as if no buttons are pressed

[garbear]

I plan to fix any and all controller issues over winter break

[bradwatson]

Any idea on when this will be merged with snapshot or nightly builds? This is the greatest xbmc project ever!

[irongecko]

Any estimate on when you are going to republish the dev branch?