Slider bodies don't show up at all on 0.10.1 on linux 64 (nouveau, GeForce GT 640)

[mossymountain]

The issue is NOT present on 0.10.0

tested with 64bit:
openjdk 1.8.0_66
openjdk 1.7.0_79
oracle java 1.8.0_51

Also, this cursor: doesn't look like itself at all on any version of opsu I've tried on linux, it's always like this (there are no artifacts in the image, it always has that square around it)
This is tested on multiple computers, including 32b ones.

A minor thing I noticed too: if a beatmap with ogg music named *.mp3 (osu plays it normally) is selected, opsu loses sound and gets very laggy, java stays 100% usage on a single core and a "Failed to destroy OpenAL" exception gets thrown every time opsu tries to load another song. This is also tested with the versions above.

.opsu.log after doing all the tests (gzipped 5415 B):
http://54.213.154.174:777/B/.opsu.log%202015-08-13%2002:45:45.gz
Opsu crashes in the end when I move the beatmap dir in question away from Songs while running with oracle java.

[itdelatrisu]

Thanks for the detailed report. I'm unfortunately not familiar enough with some of these issues to provide solutions, but maybe someone else can help.

• Is the slider issue still present if you revert Backup and restore the viewport size when rendering sliders #108? @Bigpet, do you have any insights on this?
• I don't have any ideas about the why the cursor isn't rendering correctly. x.x
• Again, not really sure what the problem is, but I think this is the relevant piece of code (written by fluddokt) in the modified org.newdawn.slick.openal.OpenALStreamPlayer.java:

            if (ref.toLowerCase().endsWith(".mp3"))
                audio = new Mp3InputStream(ResourceLoader.getResourceAsStream(ref));
            else
                audio = new OggInputStream(ResourceLoader.getResourceAsStream(ref));

            if (audio.getRate() == 0 && audio.getChannels() == 0) {
                if (ref.toLowerCase().endsWith(".mp3"))
                    audio = new OggInputStream(ResourceLoader.getResourceAsStream(ref));
                else
                    audio = new Mp3InputStream(ResourceLoader.getResourceAsStream(ref));
            }

• Good catch. I suppose I'll need to add extra file checks before trying to load any audio... but I have to think about how to actually handle cases like these.

[chong601]

also, have you tried using the proprietary nVidia driver? i think this may be related...
nouveau may not be suitable for some games because they are not really complete imo

[mad-s]

Did opsu! render the new slider style in 0.10.0 on your machine? Also what OpenGL version and GLSL version does your graphics card run? Might be related to #114 (940e9ba). According to my sources OpenGL version 3.3 is required for GLSL 330, also this fixed the issue for me, where in 0.10.0 the sliders bodies didn't render at all.
You do have some errors concerning shader compilation. Your log states that opsu! tried compiling the shaders whereas that version is clearly not supported.

[Bigpet]

@itdelatrisu nah, reverting that won't do anything. The issue is in the logs, the drivers don't support GLSL 3.30. I'll look into which version I actually need since the shaders really aren't that complex.

[mossymountain]

Tested with a machine with an i3 M380 and GeForce 310M, same results, even with nouveau blacklisted.
I'm not actually sure what drivers, or even which graphics card it uses now that it is blacklisted. There were no kernel modules loaded that were not loaded with nouveau in use.

Tested with that 310M machine with nvidia's proprietary drivers, sliders look different, this is the new style?
310M with proprietary nvidia, opsu 0.10.1:

eeepc, opsu 0.10.1, looks similar to eeepc, opsu 0.9.0
GT640, opsu 0.10.0 I can't get it to show score and the bar like the other screenshots, is there a shortcut for toggling them in opsu too?
GT640, opsu 0.10.1 and this, of course, was completely without the bodies. Strangely enough, the circle around the slider ball is visible here but not on 0.10.0.

not reproducible on an eeepc with Intel Atom N270 (i686) and its integrated graphics (945GSE)
tested with oracle java 1.8.0_25-b17

Opsu runs surprisingly well on that eeepc, only lagging noticeably when doing a long slider. (0.10.1, renders old slider style)

[itdelatrisu]

Any updates on the slider rendering after merging #122?

[mossymountain]

I'm sorry that testing this took so long! I had never compiled java before and I kinda forgot this for a while after it didn't succeed the first time I tried it.
Finally got this on the GT640 with nouveau machine by running target/opsu-0.10.1-runnable.jar with oracle's java 1.8.0_51 :

same results with openjdk 1.7.0_79 and 1.8.0_66.

This cursor thing, by the way. I only had cursor.png (like all older skins, no cursormiddle.png) in that skin and the default cursormiddle got drawn on my cursor. If I put a cursormiddle there, it replaces the default one.

Is this intended behavior not to use the skin's cursor at all with this in "No"?

Frame rate seems to stay at monitor's refresh rate plus one with nouveau if the Frame Limiter >= monitor's refresh rate (I can't get more than 61 on my LCD or more than 86 on my CRT, putting the limit of 60 on the CRT makes it stay at 61, too). Maybe this is just a thing nouveau does.

EDIT:
opsu uses the skin cursor without new style cursor toggled on if it is named cursor2.png (osu ignores it)

an example with all cursor.png, cursormiddle.png and cursor2.png present:

[itdelatrisu]

Sorry for the late reply, and thanks for providing so many details again! The cursor2/cursortrail2 was sort of a hack to begin with, and I totally forgot to account for it when implementing game skins. It should be fixed in 5efb61d -- if you use a non-default skin, the files "cursor" and "cursortrail" will always be loaded, and the "-2" files will only be used if the original files are absent and "Enable New Cursor" is disabled. (I know the last part is not a part of osu!, but I don't think it hurts.)

Also, after ae5016f, the game shouldn't crash anymore if you try to play a track that doesn't exist. The game should throw reasonable warnings if this happens in the song menu, and the music file should be locked when it's currently playing so this shouldn't happen elsewhere. I didn't test this extensively though, so let me know if there are still problems with it.

[itdelatrisu]

Okay, let me know if 20d40dd solves the remaining issue with the misnamed MP3/OGG files. I couldn't reproduce the problems you had (lag/high CPU/"Failed to destroy OpenAL") because I received null pointer exceptions immediately when playing any OGG with a .mp3 extension. If you still run into the same issue (or different ones), also tell me which beatmap it happens with. Thanks!