View Issue Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0005569Spring engineGfx driver issuespublic2017-05-22 13:352017-06-11 13:28
ReporterGoogle_Frog Assigned ToKloot  
PrioritynormalSeveritycrashReproducibilityhave not tried
Status closedResolutionfixed 
Product Version103.0 +git 
Summary0005569: 103.0.1-927 intel crash
DescriptionI am not sure whether this is reportable. Their GL version is 3.1.0 but the crash is within their graphics driver. What does the policy change mean?

https://github.com/ZeroK-RTS/CrashReports/issues/1140
TagsNo tags attached.
Checked infolog.txt for Errors

Activities

Kloot

2017-05-22 14:09

developer   ~0017673

for this player, the change won't mean much even without knowing where (in Spring) the crash originated because it didn't involve shaders.

Google_Frog

2017-05-23 11:22

reporter   ~0017675

Here are some more:
 * https://github.com/ZeroK-RTS/CrashReports/issues/1142
 * https://github.com/ZeroK-RTS/CrashReports/issues/1143

Kloot

2017-05-23 12:20

developer   ~0017677

Last edited: 2017-05-23 14:02

 1142 crashed during the feature shadow pass.

  ??:0
  rts/Rendering/Models/3DModel.cpp:240
  rts/Rendering/FeatureDrawer.cpp:390
  rts/Rendering/FeatureDrawer.cpp:333
  rts/Rendering/FeatureDrawer.cpp:553
  rts/Rendering/ShadowHandler.cpp:373
  rts/Rendering/ShadowHandler.cpp:596
  rts/Game/Game.h:108
  rts/Game/Game.cpp:1301
  rts/System/SpringApp.cpp:858
  rts/System/SpringApp.cpp:888
  rts/System/Main.cpp:46
  rts/System/Main.cpp:95

1143 is garbled beyond recognition.


HD4000 (1142) chips are barely within the GL3 era, so Shadows=0 and/or ForceDisableShaders=1 can be set to work around it. the G45 (1143) is too old and shouldn't even have gotten a valid context.

Google_Frog

2017-05-23 13:42

reporter   ~0017678

Here is another from the same source as 1142:
https://github.com/ZeroK-RTS/CrashReports/issues/1144

Kloot

2017-05-23 13:53

developer   ~0017679

terrain shadow pass this time.

  rts/Map/SMF/ROAM/RoamMeshDrawer.cpp:237
  ??:0
  ??:0
  rts/Map/SMF/SMFGroundDrawer.cpp:483
  ??:0
  rts/Rendering/ShadowHandler.cpp:384
  rts/Rendering/ShadowHandler.cpp:596
  rts/Game/Game.h:108
  rts/Game/Game.cpp:1301
  rts/System/SpringApp.cpp:858
  rts/System/SpringApp.cpp:888
  rts/System/Main.cpp:46
  rts/System/Main.cpp:95

Google_Frog

2017-05-23 16:21

reporter   ~0017680

https://github.com/ZeroK-RTS/CrashReports/issues/1145

Kloot

2017-05-23 16:34

developer   ~0017681

same trace as 1142.

ShadowMapSize = 8192 might also be beyond the HD4000's ability to handle.

Google_Frog

2017-05-27 15:24

reporter   ~0017698

Here are some more:
 * https://github.com/ZeroK-RTS/CrashReports/issues/1157
 * https://github.com/ZeroK-RTS/CrashReports/issues/1157
 * https://github.com/ZeroK-RTS/CrashReports/issues/1160
 * https://github.com/ZeroK-RTS/CrashReports/issues/1161
 * https://github.com/ZeroK-RTS/CrashReports/issues/1167
 * https://github.com/ZeroK-RTS/CrashReports/issues/1175
 * https://github.com/ZeroK-RTS/CrashReports/issues/1176

Perhaps what we know so far is the limit ShadowMapSize.

Kloot

2017-05-27 15:43

developer   ~0017700

1157-1161 are all from the same user (who seems to be trying the same settings over and over again but expecting different results), and identical to the traces above.

1167 is junk.

1175 crashed in LuaOpenGL::Texture on a glDisable(GL_TEXTURE_2D) call.

1176 crashed somewhere in LuaOpenGL::DrawListAtUnit.

at the moment I would recommend Shadows=0 and ForceDisableShaders=1.

Google_Frog

2017-05-28 03:51

reporter   ~0017718

Where do you suggest applying those settings? Would the engine detect poor graphics cards and disable shadows? Should lobbies detect crashes and suggest automatic setting adjustment?

Kloot

2017-05-28 11:42

developer   ~0017721

Last edited: 2017-05-28 11:46

 prior to the policy change the engine simply used to regard all Intels as poor and applied the equivalent of (what is now) ForceDisableShaders=1 at startup, which also disabled shadows. the recent string of Intel issues (also 1191, 1192) shows this wasn't unjustified even for GL3-era drivers, so LuaMenu lobbies could ship with an Intel settings profile and (anticipating crashes) prompt the user about enabling it on the next restart if Platform.gpuVendor == "Intel" and Platform.glVersionShort starts with a "3".

Google_Frog

2017-05-30 05:52

reporter   ~0017727

More crashes:
 * https://github.com/ZeroK-RTS/CrashReports/issues/1192
 * https://github.com/ZeroK-RTS/CrashReports/issues/1195
 * https://github.com/ZeroK-RTS/CrashReports/issues/1208
 * https://github.com/ZeroK-RTS/CrashReports/issues/1209
 * https://github.com/ZeroK-RTS/CrashReports/issues/1210

I added a compatibility profile just now. The crash history will say whether it works.

https://github.com/ZeroK-RTS/Chobby/commit/5df124c30817c2d049cf3d4e3d6de9062dcef3c8

Kloot

2017-05-30 11:51

developer   ~0017735

1192 (1142 for unit shadows)

  rts/Rendering/Models/3DModel.cpp:499
  rts/Rendering/Models/3DModel.cpp:251
  rts/Rendering/Models/3DModel.h:411
  rts/Rendering/UnitDrawer.cpp:1431
  rts/Rendering/UnitDrawer.cpp:1441
  rts/Rendering/UnitDrawer.cpp:568
  rts/Rendering/UnitDrawer.cpp:583
  rts/Rendering/UnitDrawer.cpp:619
  rts/Rendering/ShadowHandler.cpp:373
  rts/Rendering/ShadowHandler.cpp:596
  rts/Game/Game.h:108
  rts/Game/Game.cpp:1301
  rts/System/SpringApp.cpp:788
  rts/System/SpringApp.cpp:818
  rts/System/Main.cpp:46


1195 (1142 for projectile shadows)

  rts/Rendering/Env/Particles/ProjectileDrawer.cpp:700
  rts/Rendering/Env/Particles/ProjectileDrawer.cpp:468
  rts/Rendering/Env/Particles/ProjectileDrawer.cpp:454
  rts/Rendering/Env/Particles/ProjectileDrawer.cpp:636
  rts/Rendering/ShadowHandler.cpp:369
  rts/Rendering/ShadowHandler.cpp:596
  rts/Game/Game.h:108
  rts/Game/Game.cpp:1301
  rts/System/SpringApp.cpp:788
  rts/System/SpringApp.cpp:818
  rts/System/Main.cpp:46
  rts/System/Main.cpp:95


1208 is 1192.


1209
  rts/Rendering/Env/SkyBox.cpp:88
  rts/Rendering/WorldDrawer.cpp:247
  rts/Game/Game.cpp:1306
  rts/System/SpringApp.cpp:788
  rts/System/SpringApp.cpp:818
  rts/System/Main.cpp:46
  rts/System/Main.cpp:95


1210 is 1192.

Kloot

2017-06-11 13:28

developer   ~0017824

the number of crashes seems to have dropped to near-zero again.

conclusion: GL3.*-era (HD4000, etc) Intel drivers are still absolute garbage.

Issue History

Date Modified Username Field Change
2017-05-22 13:35 Google_Frog New Issue
2017-05-22 14:09 Kloot Note Added: 0017673
2017-05-23 11:22 Google_Frog Note Added: 0017675
2017-05-23 12:20 Kloot Note Added: 0017677
2017-05-23 13:42 Google_Frog Note Added: 0017678
2017-05-23 13:53 Kloot Note Added: 0017679
2017-05-23 14:02 Kloot Note Edited: 0017677
2017-05-23 16:21 Google_Frog Note Added: 0017680
2017-05-23 16:34 Kloot Note Added: 0017681
2017-05-27 15:24 Google_Frog Note Added: 0017698
2017-05-27 15:43 Kloot Note Added: 0017700
2017-05-28 03:51 Google_Frog Note Added: 0017718
2017-05-28 11:42 Kloot Note Added: 0017721
2017-05-28 11:45 Kloot Note Edited: 0017721
2017-05-28 11:46 Kloot Note Edited: 0017721
2017-05-30 05:52 Google_Frog Note Added: 0017727
2017-05-30 11:51 Kloot Note Added: 0017735
2017-06-11 13:28 Kloot Assigned To => Kloot
2017-06-11 13:28 Kloot Status new => closed
2017-06-11 13:28 Kloot Resolution open => fixed
2017-06-11 13:28 Kloot Note Added: 0017824