lr-pcsx-rearmed won't run any games

mitu

Please add more info about your system, as requested in https://retropie.org.uk/forum/topic/3/read-this-first.

What file format are your games ? Iso, Cue/Bin, PBP ?

zapp7

Pi Model or other hardware: 3B+
Power Supply used: 5V/2.5A
RetroPie Version Used (eg 3.6, 3.8.1, 4.1 - do not write latest): 4.4
Built From: installed with install-retrosmc script on OSMC
USB Devices connected: 4TB WD hard drive
Controller used: PS4 controller
Error messages received: game won't start
Log found in /dev/shm/runcommand.log (if relevant):
Guide used: n/a
File: psx games located in /home/osmc/RetroPie/roms/psx
Emulator: lr-pcsx-rearmed

Edit: games are in Cue/Bin format

mitu

@zapp7 Check the .cue file contents (it's a text file) and make sure the .bin filenames in the file match the actual .bin file names on disk. Linux is case-sensitive on file names, so they have to match exactly.
If that doesn't work, then enable verbose logging when starting a game - from the Runcommand launch menu - and then get put the log file (/dev/shm/runcommand.log) on pastebin.com so we can take a look.

zapp7

Here is the log:

Parameters: Executing: /opt/retropie/emulators/retroarch/bin/retroarch
-L /opt/retropie/libretrocores/lr-pcsx-rearmed/pcsx_rearmed_libretro.so
--config /opt/retropie/configs/psx/retroarch.cfg
"/home/osmc/RetroPie/roms/psx/Resident Evil 3 - Nemesis (USA).cue"
--verbose --appendconfig /dev/shm/retroarch.cfg [INFO] RetroArch 1.7.6
(Git 9750719) [INFO] Redirecting save file to
"/home/osmc/RetroPie/roms/psx/Resident Evil 3 - Nemesis (USA).srm".
[INFO] Redirecting savestate to "/home/osmc/RetroPie/roms/psx/Resident
Evil 3 - Nemesis (USA).state". [INFO] === Build
======================================= Capabilities: NEON VFPv3 VFPv4
Built: Feb 4 2019 [INFO] Version: 1.7.6 [INFO] Git: 9750719 [INFO]
================================================= [ERROR] RetroArch is
built for dynamic libretro cores, but libretro_path is not set. Cannot
continue. [ERROR] Fatal error received in: "init_libretro_sym()"

mitu

@zapp7 Looks like the lr-pcsx_rearmed emulator is not installed - does the /opt/retropie/libretrocores/lr-pcsx-rearmed/pcsx_rearmed_libretro.so file exist ?

zapp7

The only .so file in that directory is libretro.so.

mitu

Then your installation is incomplete.

zapp7

Oh okay. Is there a resource that you recommend to install RetroPie completely? I was under the impression that the setup script would take care of that.

zapp7

I used the update script and it already said that lr-pcsx-rearmed was installed. I chose "Update from Source" anyway and now it's working!

Thank you for your help!

hhromic

@mitu @BuZz due to the recent lr-pcsx-rearmed buildfix that now uses the libretro makefile directly, the generated shared file is now called pcsx_rearmed_libretro.so, which the scriptmodule configures in emulators.cfg. The old name was just libretro.so.

I think the problem here is that the curent binaries in the RetroPie archive probably are the old ones installing the libretro.so and the new scriptmodule is configuring for pcsx_rearmed_libretro.so. This is why installing from source works. The solution should be simply to update the binaries on the archive server.

EDIT: apologies as I forgot to remind in the PR to update the binaries.

BuZz

@hhromic I will sort. Sorry, just been a bit busy. Rebuilding now.

BuZz

Build is currently failing for RPI1. Looks like neon code is trying to be built. I won't be able to debug the RPI1 build till later.

RPI2 bins are updated.

hhromic

@BuZz umm I tested that build too, will take another closer look and report back. Hopefully I find the culprit before you need to spend time on it.

BuZz

@hhromic possibly due to logic somewhere that would fail then if built on my chroot. Eg. Something that checks for RPI1 hardware info. I have a feeling there was something before we had as a workaround for a similar issue.

Actually that's for the standalone core. Maybe it's the same issue though.

josephxx

Hello world (first post by a new owner of a RPI 3B+):

To anyone who (like me) ended up here having the same problem/symptoms, the reason in my case was that I allowed my 4.4 Retropie to update everything.

Stock 4.4 image played PSX roms perfectly. When I read that a recent update to EmulationStation allowed the internal scraper to start working again, I did a full update. The scraper indeed worked, but lr-pcsx-rearmed never launched another ROM - black screen, starting, and back to the game list again.

What worked was updating specifically the EmulationStation package and nothing else.

mitu

@josephxx You can update from source the emulator and it should work.

hhromic

@BuZz said in lr-pcsx-rearmed won't run any games:

@hhromic possibly due to logic somewhere that would fail then if built on my chroot. Eg. Something that checks for RPI1 hardware info. I have a feeling there was something before we had as a workaround for a similar issue.

Actually that's for the standalone core. Maybe it's the same issue though.

Thanks for the info, will investigate that as I was having the same feeling. Libretro makefiles can be a bit hacky sometimes. Yes I think I saw the workaround you mention in the standalone emulator, however I will check passing variables to the makefile, e.g. HAVE_NEON=0, if necessary. I will be able to report shortly.

josephxx

@mitu said in lr-pcsx-rearmed won't run any games:

@josephxx You can update from source the emulator and it should work.

Thanks for letting me know. I am new to all this and the Linux mindset (and it's been 20 years since I took some Linux at uni).

hhromic

@buzz the build for RPI1 is now failing here for me as well doing the same i did before, so something changed upstream since I made the buildfix.

I will investigate what they changed and send a patch upstream as I believe our way to build is correct and we shouldn't use workarounds. They are rather quick to merge PRs for this core anyway. Will report back here anyway to keep you updated guys.

hhromic

Okay found the problem.

It is upstream but nothing they did recently actually. As you guessed @BuZz, when HAVE_NEON is not set by any of the checks the makefile tries to autodetect it using a compiler test. The problem is that when platform=armvalone, the makefile starts with HAVE_NEON unset, however for armv if neon is not set, then it will try to autodetect it using the compiler of the host in the chroot, leaving it enabled. The solution is rather simply to not leave the variable unset in the armv check block, i.e. set HAVE_NEON=0 explicitly in armv until NEON is requested explicitly. This way there is no autodetection, like this:

index 8899ed7..c6eb695 100644
--- a/Makefile.libretro
+++ b/Makefile.libretro
@@ -286,6 +286,7 @@ else ifneq (,$(findstring armv,$(platform)))
        TARGET := $(TARGET_NAME)_libretro.so
        fpic := -fPIC
        DRC_CACHE_BASE = 0
+       HAVE_NEON = 0
        ifneq (,$(findstring cortexa8,$(platform)))
                CFLAGS += -marm -mcpu=cortex-a8
                ASFLAGS += -mcpu=cortex-a8

If NEON is present in platform=armvneon then HAVE_NEON=1 will be set correctly.

I will send a patch upstream now. An alternative is to pass HAVE_NEON accordingly from the scriptmodule but that will make the logic unnecessarily more complicated when it can be fixed upstream very easily.

I don't know exactly why my initial test worked, perhaps the compiler check did work at some point. Will let you know when the patch is merged upstream so you can rebuild.