Artifacts and other things in 1.79.0.5527-140112-8d3ef6e

[Diziet Sma]

Oolite development version 1.79.0.5527-140112-8d3ef6e (x86-64 test release - Linux)

It's been a while since I've seen anything like this.. so I'm wondering if it's somehow related to the re-enabling of New Planets (which is looking good, btw, apart from that sharp-as-a-knife day/night terminator line).

I launched from a station, did my usual full-throttle and slow pitch up until I was heading directly away from the planet, and happened to check the rear view as I was swinging around to my departure heading. Loads of very rapidly flickering jagged artifacts apparently coming from the vicinity of my port engine. Throttle setting changes had no effect on it. No other views showed it, and after some manoeuvring and view changes, it was gone.. I got 7 screenshots of it in quick succession. Here's one example:



When I got to my destination station, I found a Sidewinder hanging motionless in front of the docking bay.. several other ships were hanging around, waiting for this guy to make up his mind what he wanted to do.. after a few minutes it became clear he was in no hurry to move, so I targeted him and did a dumpState, then docked to make this report.

It appears he exited the station, then parked himself about a kilometre outside the docking bay and just sat there twiddling his thumbs.. he seems to have been shot once, about a second after spawning.. maybe the experience left him in a blue funk..

Here's the relevant section of the log:

Code: Select all

08:00:08.438 [dumpState]: State for <ShipEntity 0x82cdc60>{"Sidewinder Scout Ship" position: (71595, -97240.1, 664264) scanClass: CLASS_NEUTRAL status: STATUS_IN_FLIGHT}:
  08:00:08.439 [dumpState.entity]: Universal ID: 621
  08:00:08.439 [dumpState.entity]: Scan class: CLASS_NEUTRAL
  08:00:08.439 [dumpState.entity]: Status: STATUS_IN_FLIGHT
  08:00:08.439 [dumpState.entity]: Position: (71595, -97240.1, 664264)
  08:00:08.439 [dumpState.entity]: Orientation: (0.615384 + 0.72151i + 0.218658j - 0.230031k)
  08:00:08.439 [dumpState.entity]: Distance travelled: 961.689
  08:00:08.439 [dumpState.entity]: Energy: 240 of 240
  08:00:08.439 [dumpState.entity]: Mass: 25087.6
  08:00:08.439 [dumpState.entity]: Owner: self
  08:00:08.439 [dumpState.entity]: Flags: isShip, isSunlit
  08:00:08.439 [dumpState.entity]: Collision Test Filter: 0
  08:00:08.440 [dumpState.shipEntity]: Type: sidewinder
  08:00:08.440 [dumpState.shipEntity]: Name: Sidewinder Scout Ship
  08:00:08.440 [dumpState.shipEntity]: Display Name: Sidewinder Scout Ship: The Hydra, Enchanted
  08:00:08.440 [dumpState.shipEntity]: Roles: <OORoleSet 0xf4392f0>{[sidewinder] hunter pirate(0.75) pirate-light-fighter(0.25)}
  08:00:08.440 [dumpState.shipEntity]: Primary role: hunter
  08:00:08.440 [dumpState.shipEntity]: Script: <OOJSScript 0x3b90350>{"oolite-default-ship-script" version 1.79}
  08:00:08.440 [dumpState.shipEntity]: Subentity count: 1
  08:00:08.440 [dumpState.shipEntity]: Behaviour: BEHAVIOUR_FLY_TO_DESTINATION
  08:00:08.440 [dumpState.shipEntity]: Target: <none>
  08:00:08.440 [dumpState.shipEntity]: Destination: (66512.7, -90534.9, 663087)
  08:00:08.440 [dumpState.shipEntity]: Other destination: (0, 0, 0)
  08:00:08.440 [dumpState.shipEntity]: Waypoint count: 0
  08:00:08.440 [dumpState.shipEntity]: Desired speed: 0
  08:00:08.441 [dumpState.shipEntity]: Thrust: 37
  08:00:08.441 [dumpState.shipEntity]: Fuel: 0
  08:00:08.441 [dumpState.shipEntity]: Fuel accumulator: 1
  08:00:08.441 [dumpState.shipEntity]: Missile count: 0
  08:00:08.441 [dumpState.shipEntity.ai]: AI:
    08:00:08.441 [dumpState.ai]: State machine name: nullAI.plist
    08:00:08.441 [dumpState.ai]: Current state: GLOBAL
    08:00:08.441 [dumpState.ai]: Next think time: 4291.53
    08:00:08.441 [dumpState.ai]: Next think interval: 0.125
  08:00:08.441 [dumpState.shipEntity]: Jink position: (0, 0, 0)
  08:00:08.441 [dumpState.shipEntity]: Frustration: 10.4705
  08:00:08.441 [dumpState.shipEntity]: Success factor: 8495.58
  08:00:08.441 [dumpState.shipEntity]: Shots fired: 0
  08:00:08.441 [dumpState.shipEntity]: Time since shot: 683.693
  08:00:08.441 [dumpState.shipEntity]: Spawn time: 587.677 (684.852 seconds ago)
  08:00:08.442 [dumpState.shipEntity]: Hull temperature: 60
  08:00:08.442 [dumpState.shipEntity]: Heat insulation: 1
  08:00:08.442 [dumpState.shipEntity]: Flags: isFrangible, canFragment
08:02:05.719 [exit.context]: Exiting: Exit Game selected on options screen.

[another_commander]

Oolite development version 1.79.0.5527-140112-8d3ef6e (x86-64 test release - Linux)

...related to the re-enabling of New Planets (which is looking good, btw, apart from that sharp-as-a-knife day/night terminator line).

Not sure about the Sidewinder AI issue, but I am very interested in the quoted part, as I am getting the exact same thing on Windows. The terminator should be smooth, but for some reason on some configurations it apparently isn't. Can you please post a screen of what it looks like and your gfx card/driver specs?

For reference, here is what it looks like for me on an Intel HD Graphics with driver version 8.15.10.2418:

[Cody]

This is how I'm seeing the terminator: close-up, and at distance (that's Estiri, btw).

[another_commander]

This is how I'm seeing the terminator: close-up, and at distance (that's Estiri, btw).

Yes, this is what you are supposed to see.

[Cody]

Yes, this is what you are supposed to see.

As I thought, yes... but no atmosphere effect?

[cim]

Yes, this is what you are supposed to see.

As I thought, yes... but no atmosphere effect?

There's definitely an atmosphere effect on both of those. It's probably most obvious near the poles on the long-distance picture, I think.

[Cody]

There's definitely an atmosphere effect on both of those.

<cleans glasses> Hmm... so does the effect vary from planet to planet? It's much clearer here, for example.

The planets are looking really good, btw.



Apologies to Diz for the slight hijack.

[cim]

Yes. Side-on you can only really see it where there's thick cloud. Estiri has black clouds, so you can see it quite easily against the planet - around the poles, and above the ocean to the left - but it really doesn't show up much against space. The amount of cloud cover varies from planet to planet, too.

[Diziet Sma]

I am very interested in the quoted part, as I am getting the exact same thing on Windows. The terminator should be smooth, but for some reason on some configurations it apparently isn't. Can you please post a screen of what it looks like and your gfx card/driver specs?

Sure.. here's a screenshot:




And here are my graphics specs:

Code: Select all

16:35:51.228 [display.mode.list.native]: X11 native resolution detected: 1280 x 800
16:35:51.482 [joystick.init]: Number of joysticks detected: 1
16:35:51.483 [rendering.opengl.version]: OpenGL renderer version: 2.1.0 ("2.1 Mesa 9.2.1"). Vendor: "X.Org R300 Project". Renderer: "Gallium 0.4 on ATI RV515".
16:35:51.483 [rendering.opengl.extensions]: OpenGL extensions (141):
GL_ATI_texture_mirror_once, GL_ARB_depth_texture, GL_ARB_invalidate_subdata, GL_ARB_texture_env_add, GL_NV_texture_barrier, GL_EXT_pixel_buffer_object, GL_EXT_texture_compression_rgtc, GL_ARB_half_float_vertex, GL_IBM_multimode_draw_arrays, GL_EXT_texture3D, GL_EXT_compiled_vertex_array, GL_ARB_framebuffer_object, GL_EXT_separate_specular_color, GL_EXT_shadow_funcs, GL_ARB_shader_texture_lod, GL_EXT_texture_cube_map, GL_EXT_rescale_normal, GL_ATI_draw_buffers, GL_ARB_transpose_matrix, GL_ARB_texture_compression_rgtc, GL_ARB_half_float_pixel, GL_EXT_stencil_two_side, GL_ARB_texture_border_clamp, GL_ARB_multisample, GL_ATI_separate_stencil, GL_ARB_texture_non_power_of_two, GL_NV_fog_distance, GL_INGR_blend_func_separate, GL_EXT_bgra, GL_ARB_sync, GL_EXT_texture_env_dot3, GL_EXT_provoking_vertex, GL_APPLE_packed_pixels, GL_NV_light_max_exponent, GL_ARB_sampler_objects, GL_SUN_multi_draw_arrays, GL_ARB_map_buffer_range, GL_ARB_vertex_shader, GL_ARB_shader_objects, GL_NV_texgen_reflection, GL_EXT_multi_draw_arrays, GL_ARB_vertex_program, GL_IBM_rasterpos_clip, GL_EXT_texture_object, GL_ARB_get_program_binary, GL_ATI_texture_env_combine3, GL_ARB_texture_float, GL_NV_texture_env_combine4, GL_ATI_texture_compression_3dc, GL_ARB_fragment_coord_conventions, GL_EXT_stencil_wrap, GL_EXT_texture_swizzle, GL_EXT_framebuffer_blit, GL_NV_primitive_restart, GL_MESA_pack_invert, GL_ARB_vertex_array_bgra, GL_NV_packed_depth_stencil, GL_ARB_texture_compression, GL_ARB_vertex_array_object, GL_EXT_blend_color, GL_EXT_blend_minmax, GL_ARB_texture_swizzle, GL_APPLE_vertex_array_object, GL_EXT_subtexture, GL_ARB_occlusion_query2, GL_MESA_texture_signed_rgba, GL_ARB_debug_output, GL_ARB_point_sprite, GL_EXT_vertex_array, GL_EXT_abgr, GL_ARB_vertex_buffer_object, GL_NV_blend_square, GL_ARB_instanced_arrays, GL_EXT_framebuffer_multisample, GL_NV_conditional_render, GL_NV_texture_rectangle, GL_ARB_multitexture, GL_EXT_texture, GL_EXT_vertex_array_bgra, GL_ARB_copy_buffer, GL_ARB_draw_elements_base_vertex, GL_ARB_shadow, GL_EXT_blend_subtract, GL_ARB_internalformat_query, GL_ARB_fragment_program_shadow, GL_MESA_window_pos, GL_ARB_texture_storage, GL_ARB_window_pos, GL_EXT_texture_lod_bias, GL_IBM_texture_mirrored_repeat, GL_EXT_texture_snorm, GL_ARB_texture_env_crossbar, GL_ARB_point_parameters, GL_EXT_texture_sRGB_decode, GL_ARB_robustness, GL_ARB_texture_rg, GL_EXT_texture_edge_clamp, GL_EXT_copy_texture, GL_EXT_texture_sRGB, GL_MESA_ycbcr_texture, GL_ARB_texture_env_combine, GL_ARB_fragment_program, GL_ARB_explicit_attrib_location, GL_EXT_texture_filter_anisotropic, GL_EXT_draw_range_elements, GL_ARB_pixel_buffer_object, GL_EXT_packed_depth_stencil, GL_ARB_texture_rectangle, GL_EXT_fog_coord, GL_ARB_ES2_compatibility, GL_SGIS_texture_lod, GL_EXT_gpu_program_parameters, GL_ARB_fragment_shader, GL_OES_EGL_image, GL_EXT_polygon_offset, GL_ARB_shading_language_100, GL_SGIS_texture_border_clamp, GL_EXT_secondary_color, GL_EXT_point_parameters, GL_ARB_draw_buffers, GL_SGIS_generate_mipmap, GL_EXT_texture_rectangle, GL_ATI_blend_equation_separate, GL_OES_read_format, GL_EXT_blend_func_separate, GL_ARB_texture_cube_map, GL_ARB_occlusion_query, GL_EXT_texture_mirror_clamp, GL_EXT_framebuffer_object, GL_EXT_packed_pixels, GL_ARB_texture_mirrored_repeat, GL_ARB_texture_env_dot3, GL_ARB_map_buffer_alignment, GL_EXT_texture_compression_latc, GL_ARB_provoking_vertex, GL_EXT_blend_equation_separate, GL_ARB_color_buffer_float, GL_ATI_texture_float, GL_EXT_texture_env_add, GL_EXT_texture_env_combine, GL_SGIS_texture_edge_clamp
16:35:51.728 [rendering.opengl.shader.support]: Shaders are supported.

Another shot of the artifacts, which I've found I can consistently generate in the most-planetwards view, whenever the planet is at roughly 90° to the orientation of the ship.. ie, when the planet is at the bottom or top of the ASC. This leads me to think it is indeed somehow related to the planetary rendering.

[another_commander]

I have OpenGL version 2.1.0 as well. I wonder whether this has to do with the planet render issue.

On a separate note, I have received a few CTDs while cruising normally from witchspace to planet. I'll try to run the game under gdb for a while and see if I can grab a backtrace.

[cim]

If you turn the graphics settings down from maximum to return to something similar to the old rendering, do you still get those artifacts? If so, what about if you turn them right down to the minimum?

OpenGL 2.1.2 here, so it's possible that this makes a difference.

[another_commander]

If you turn the graphics settings down from maximum to return to something similar to the old rendering, do you still get those artifacts? If so, what about if you turn them right down to the minimum?

OpenGL 2.1.2 here, so it's possible that this makes a difference.

In my case, going to detail normal or no shaders will render the planets correctly, although their look is closer to the old style.

[Diziet Sma]

If you turn the graphics settings down from maximum to return to something similar to the old rendering, do you still get those artifacts? If so, what about if you turn them right down to the minimum?

On Normal or Minimum, the terminator line looks better, but the artifacts remain in both cases.

(A minor point.. when changing the graphics settings, right-arrow will stop at Extra, as it should, but if I toggle left-arrow, I can go right past Minimum to Extra again, and continue cycling through the three settings in that direction as much as I want.)

(Another thought.. it would be nice if, when changing a particular setting, you got a confirming beep, the same [or maybe with a different tone] as when you scroll up and down through the list.)

[cim]

If you turn the graphics settings down from maximum to return to something similar to the old rendering, do you still get those artifacts? If so, what about if you turn them right down to the minimum?

OpenGL 2.1.2 here, so it's possible that this makes a difference.

In my case, going to detail normal or no shaders will render the planets correctly, although their look is closer to the old style.

...and, looking it up, smoothstep was only standardised in GLSL 1.3 (OpenGL 3.0), so it's quite likely the 2.x series cards only have a partial (or perhaps no) implementation.

I have GL_ARB_shading_language_420pack in my OpenGL extensions, and Diziet Sma's log doesn't. That one apparently requires GLSL 1.3, so we can probably test for that and switch to using the internal smoothing interpretation if it's not present.

(Though I suspect, in the long run, we'll probably end up with 3.0-class hardware being needed to run "Extras" sensibly, at the moment that's not quite the case)

EDIT: Oh, another thing to try with the artifacts: do they still appear (at any/all graphics settings) if you install a planet texture pack? (Try both a cube-map one like Povray and a latlong one like System Redux, please)

[another_commander]

Regarding smoothstep, when I was trying to debug the planet fragment shader, I ended up also testing a smoothstep implementation I found at an Nvidia development site, which was basically this:

Code: Select all

float saturate(float x)
{
  return max(0, min(1, x));
}


float smoothstep(float a, float b, float x)
{
    float t = saturate((x - a)/(b - a));
    return t*t*(3.0 - (2.0*t)); 
}

No idea if this is supposed to be the official implementation, but I tried using this code on the shader instead of the built-in smoothstep and did not see any improvement in the situation. I am not excluding the possibility of making a mistake somewhere in the shader while doing this, of course.