Citra: Ever Oasis - freezes when creating a save file

[GDhex12]

Using: Citra Nightly, hash - cac1133 .

The game freezes up when creating a new save file. The screen turns white and throws unmapped Read32 errors.

How to reproduce: Load up the game, press A and then try to create a save file.

Specs: Intel Core i3-6100, Nvidia GeForce GTX 750 Ti, 6 GB RAM

Log file right here: https://pastebin.com/9snDKWs1

[yuukiw]

tried adding a fresh system archive in case and tried several saves from a 3ds.
With a save it will always crash when loading said saves.

https://pastebin.com/YdkjVFVU

[Marborao]

Any progress? Can someone give some love to this game? thank you.

[ghost]

New update from me drwhojan here, interesting..
https://gbatemp.net/threads/citra-unofficial-chinese-builds-discussion.431974/page-310#post-7658005

Core timing are "microseconds" Event ?

EDIT: Although if you close the emulator down just after starting a new save, and reset the save, speed triples.

[ghost]

Edit: seems to be some sort of Down Count.

[ghost]

@Hexagon12 test with #3091

[FearlessTobi]

Just tested it. The freeze doesn´t seem to be fixed.

[ghost]

@FearlessTobi game region and citra region?

[FearlessTobi]

Europe and Europe. But why should that be important in any way?

[ghost]

freeze with europe/usa/jpn and auto, europe and usa, europe and jpn

[bakugo]

Using the latest commit https://github.com/citra-emu/citra/pull/3091/commits/df7808c9e134343ada3e31c664759757e563f1fb from https://github.com/citra-emu/citra/pull/3091 fixes this freeze.

I tried loading the save from my 3ds to see how it works ingame: the screen is entirely black, but I can hear the game running in the background and my character moving when I press buttons. I'll test it some more later today.

[PPLToast]

Can confirm here it works too, though yeah you have to pause and unpause the game to see the screen. Strange bug.

[PananaPhone]

It appear the game freeze again with same error after long unskippable intro.

[bakugo]

It only freezes if you had an existing save and try to start the game with it. Delete the save before starting and it should work. Protip: disable vsync and framerate limit to get past the cutscene pretty fast.

Though it looks like many things don't trigger properly after the game starts (when you first go ingame a conversation should trigger before you can walk around, but that doesn't happen, and other things needed to continue don't work either) so it's impossible to progress.

[B3n30]

Closed by #3091

[bakugo]

This is not fixed in the current nightly, please re-open. Only https://github.com/citra-emu/citra/commit/df7808c9e134343ada3e31c664759757e563f1fb worked.

[Senjosei]

Tested on canary 268
If u set the IPCDelayNanoseconds to 10000000 it loads correctly with all the triggers but the game runs super slow so i debug outputed the thread names and theres this weird Path: [Binary: 000000000000000000000000]_Server . If u only delay the Path: [Binary: 000000000000000000000000]_Server it runs smoothly without any errors

if(name == "Path: [Binary: 000000000000000000000000]_Server"){
    static constexpr u64 IPCDelayNanoseconds = 10000000;
    thread->WakeAfterDelay(IPCDelayNanoseconds);
} else {
    static constexpr u64 IPCDelayNanoseconds = 51140;
    thread->WakeAfterDelay(IPCDelayNanoseconds);
}

*Edit: Noob assumption (cuz im still trying to understand how 3ds&citra works) for sync requests involving the file system, the thread should be delayed proportionally with the file size ?

[bakugo]

Before I continue, let me start by saying that I have no idea what I'm doing.

I did some testing, and the only ServerSession that needs to be delayed is one created by Service::FS::File::Connect when the game opens its own RomFS (it has the name Path: [Binary: 000000000000000000000000]_Server as mentioned by @Senjosei). No others need to be delayed for the game to load properly in all cases I tested. The opened file is 890961920 bytes and the minimum delay I was able to use was 1500000 nanoseconds.

This is probably a stupid question, but isn't this just caused by file read delays being longer on hardware than on Citra?

[B3n30]

@bakugo Indeed thats the issue. Same is probably true for some other file-IOs. Our plan is to "measure" how long HW takes for that operations and add a proper delay to just those functions.

[wwylele]

Fixed by #3440