Description: When I start Libreoffice 7.0 in my Windows 10, it shows garbled icons, like corrupt. I have skia/Vulkan acceleration enabled and latest graphics drivers installed (Intel HD 530). Steps to Reproduce: 1.Switch on Libreoffice 2. 3. Actual Results: Icons appear inexistent or garbled/corrupt, although I can toggle the icons in question. Expected Results: Icons appear normal Reproducible: Sometimes User Profile Reset: Yes OpenGL enabled: Yes Additional Info: Version: 7.0.0.3 (x64) Build ID: 8061b3e9204bef6b321a21033174034a5e2ea88e Subprocs. CPU: 4; SO: Windows 10.0 Build 19041; Repres. IU: Skia/Vulkan; VCL: win Locale: es-ES (es_ES); Interfaz: es-ES Calc: threaded
Created attachment 164584 [details] Icons garbled on Windows
This is only with Skia?
Yes, this is only with Skia enabled.
Please see this comment on how to provide your skia.log: https://bugs.documentfoundation.org/show_bug.cgi?id=129062#c4
Ok, I will provide Skia logs here. Thanks.
I assume Vulkan only? Or also with Raster (Software mode)
Created attachment 164586 [details] Here's the Skia log Well. I have found a strange behaviour. The skia log points that the discrete graphics (Nvidia GTX960M) are used in Writer when I use the program. But, when I try to change the card from Nvidia to Intel in Nvidia Control Panel, it happens the garbled icons.
(In reply to Telesto from comment #6) > I assume Vulkan only? Or also with Raster (Software mode) Yes. It happens only on Vulkan/Skia mode.
Please post result of msinfo32.exe run. Need the top of the Summary, and content the Components -> Display panel. Seems you are on a laptop so your choice to bypass vendor support and force on more recent drivers. Current Intel DCH driver for Windows 10 is 27.20.100.8587 https://downloadcenter.intel.com/product/88345/Intel-HD-Graphics-530 And while the Vulkan log you posted shows nVidia GTX 960M driver reporting Vulkan 1.2 support--not clear it is a full compliment. Reported 451 series Vulkan driver for the nVidia 960M looks to be reasonably current (can't tell release directly), 452.06 is the latest WHQL cert DCH line. Maybe a run of of the Windows build of Realtech's OpenGL Extensions Viewer which now includes Vulkan testing. https://www.realtech-vr.com/home/glview A quick check of nVidia forums some with user issues controlling nVidia Optimus application selection between GPUs on Windows 10 builds. Are you able to change the Skia rendering onto the Intel HD Graphics 530? Which GPU shows active in the Skia.log when you select Tools -> Options -> View and 'Force Skia software rendering'--if not, ending up with Default GDI rendering, could have a driver denylist issue--opengl_denylist_windows.xml. Details from the msinfo32.exe panels will help tease that out.
Created attachment 164598 [details] msinfo32.exe capture Well, that's the msinfo32.exe information
Created attachment 164599 [details] That's the skia log when Skia is forced
Created attachment 164600 [details] Skia forced by software screen Well, that's a curious thing. When I force the program to represent the windows with Skia, it shows me a software renderer. The Intel and Nvidia drivers has been upgraded to the latest version available. Bug of the Optimus system, I guess?
Thanks, but also need to 'Componentes' -- monitor or display details. Text from the panels can be copied and pasted into the comment--no need to screen clip.
In "Componentes->Pantalla": Nombre Intel(R) HD Graphics 530 Id. de dispositivo PNP PCI\VEN_8086&DEV_191B&SUBSYS_1C5D1043&REV_06\3&11583659&0&10 Tipo de adaptador Intel(R) HD Graphics Family, compatible con Intel Corporation Descripción de adaptador Intel(R) HD Graphics 530 RAM de adaptador 1,00 GB (1.073.741.824 bytes) Controladores instalados C:\WINDOWS\System32\DriverStore\FileRepository\iigd_dch.inf_amd64_9d53c3023a3ba6b6\igdumdim64.dll,C:\WINDOWS\System32\DriverStore\FileRepository\iigd_dch.inf_amd64_9d53c3023a3ba6b6\igd10iumd64.dll,C:\WINDOWS\System32\DriverStore\FileRepository\iigd_dch.inf_amd64_9d53c3023a3ba6b6\igd10iumd64.dll,C:\WINDOWS\System32\DriverStore\FileRepository\iigd_dch.inf_amd64_9d53c3023a3ba6b6\igd12umd64.dll Versión del controlador 27.20.100.8587 Archivo INF oem46.inf (sección iSKLD_w10_DS) Esquema de color No disponible Entradas de tabla de colores 4294967296 Resolución 1920 x 1080 x 60 hercios Bits/píxel 32 Dirección de memoria 0xDD000000-0xDDFFFFFF Dirección de memoria 0xB0000000-0xBFFFFFFF Puerto E/S 0x0000F000-0x0000F03F Canal IRQ IRQ 4294967285 Puerto E/S 0x000003B0-0x000003BB Puerto E/S 0x000003C0-0x000003DF Dirección de memoria 0xA0000-0xBFFFF Controlador C:\WINDOWS\SYSTEM32\DRIVERSTORE\FILEREPOSITORY\IIGD_DCH.INF_AMD64_9D53C3023A3BA6B6\IGDKMD64.SYS (27.20.100.8587, 25,95 MB (27.210.592 bytes), 18/08/2020 0:21) Nombre NVIDIA GeForce GTX 960M Id. de dispositivo PNP PCI\VEN_10DE&DEV_139B&SUBSYS_1C5D1043&REV_A2\4&1EAFE439&0&0008 Tipo de adaptador GeForce GTX 960M, compatible con NVIDIA Descripción de adaptador NVIDIA GeForce GTX 960M RAM de adaptador (2.147.483.648) bytes Controladores instalados C:\WINDOWS\System32\DriverStore\FileRepository\nvami.inf_amd64_326005dc610b0539\nvldumdx.dll,C:\WINDOWS\System32\DriverStore\FileRepository\nvami.inf_amd64_326005dc610b0539\nvldumdx.dll,C:\WINDOWS\System32\DriverStore\FileRepository\nvami.inf_amd64_326005dc610b0539\nvldumdx.dll,C:\WINDOWS\System32\DriverStore\FileRepository\nvami.inf_amd64_326005dc610b0539\nvldumdx.dll Versión del controlador 27.21.14.5206 Archivo INF oem58.inf (sección Section009) Esquema de color No disponible Entradas de tabla de colores No disponible Resolución No disponible Bits/píxel No disponible Dirección de memoria 0xDE000000-0xDF0FFFFF Dirección de memoria 0xC0000000-0xD1FFFFFF Dirección de memoria 0xD0000000-0xD1FFFFFF Canal IRQ IRQ 4294967286 Controlador C:\WINDOWS\SYSTEM32\DRIVERSTORE\FILEREPOSITORY\NVAMI.INF_AMD64_326005DC610B0539\NVLDDMKM.SYS (27.21.14.5206, 23,55 MB (24.694.504 bytes), 23/08/2020 16:20)
(In reply to impreza233 from comment #12) > Well, that's a curious thing. When I force the program to represent the > windows with Skia, it shows me a software renderer. The Intel and Nvidia > drivers has been upgraded to the latest version available. Bug of the > Optimus system, I guess? And, are the garbled/corrupt icons still present in Skia 'software' raster rendering?
(In reply to V Stuart Foote from comment #15) > (In reply to impreza233 from comment #12) > > > Well, that's a curious thing. When I force the program to represent the > > windows with Skia, it shows me a software renderer. The Intel and Nvidia > > drivers has been upgraded to the latest version available. Bug of the > > Optimus system, I guess? > > And, are the garbled/corrupt icons still present in Skia 'software' raster > rendering? No. Icons are alive and in correct shape in that mode.
Hmm, OK. So broken Vulkan support on that GPU with the 452.06 driver (Vulkan 451.xx) on Windows 10 (2004). I'll check on a Windows 10 box (1909) with same nVidia Maxwell 1 architecture as your 960M (a GTX 750 TI), and another with earlier Kepler architecture (GTX 730). And maybe for giggles throw together a Pascal architecture setup (GTX 1030). But if you want to venture out onto the bleeding edge--you could check for behavior with current nVidia BETA 451.98 Vulkan developmental drivers. https://developer.nvidia.com/vulkan-driver Changes there eventually make it into the release GeForce drivers so you'd know if there was any hope for your GPU. Meanwhile, may require a deny_list entry for the hardware/driver pair to suppress Skia Vulkan rendering: Vendor: 0x10de Device: 0x139b Driver: 451.268.0
(In reply to V Stuart Foote from comment #17) > Hmm, OK. So broken Vulkan support on that GPU with the 452.06 driver (Vulkan > 451.xx) on Windows 10 (2004). > > I'll check on a Windows 10 box (1909) with same nVidia Maxwell 1 > architecture as your 960M (a GTX 750 TI), and another with earlier Kepler > architecture (GTX 730). And maybe for giggles throw together a Pascal > architecture setup (GTX 1030). > > But if you want to venture out onto the bleeding edge--you could check for > behavior with current nVidia BETA 451.98 Vulkan developmental drivers. > https://developer.nvidia.com/vulkan-driver > > Changes there eventually make it into the release GeForce drivers so you'd > know if there was any hope for your GPU. > > Meanwhile, may require a deny_list entry for the hardware/driver pair to > suppress Skia Vulkan rendering: > Vendor: 0x10de > Device: 0x139b > Driver: 451.268.0 Ok, I will check and comment about that. Thanks!! I support the deny list entry. The garbled icons are very unpleasant and can cause problems on users.
Luboš Luňák committed a patch related to this issue. It has been pushed to "master": https://git.libreoffice.org/core/commit/f8d6eb6f171dbcf7e3aa4afa0da32bd504f473dc blacklist Windows/nVidia for Skia/Vulkan (tdf#135370, tdf#136036, tdf#136407) It will be available in 7.1.0. The patch should be included in the daily builds available at https://dev-builds.libreoffice.org/daily/ in the next 24-48 hours. More information about daily builds can be found at: https://wiki.documentfoundation.org/Testing_Daily_Builds Affected users are encouraged to test the fix and report feedback.
Cannot reproduce with Windows 10 GeForce GTX 1070 Driver version 451.67
Luboš Luňák committed a patch related to this issue. It has been pushed to "libreoffice-7-0": https://git.libreoffice.org/core/commit/3f7e5e0fa3916cfb80212969257bee58dae96c54 blacklist Windows/nVidia for Skia/Vulkan (tdf#135370, tdf#136036, tdf#136407) It will be available in 7.0.3. The patch should be included in the daily builds available at https://dev-builds.libreoffice.org/daily/ in the next 24-48 hours. More information about daily builds can be found at: https://wiki.documentfoundation.org/Testing_Daily_Builds Affected users are encouraged to test the fix and report feedback.
@impreza233, is this issue still reproducible in LibreOffice 7.0.3 ?
No. I can't reproduce this behaviour on the current version, and in recent Windows 10 20H2.
All but most recent nvidia drivers have been blocked, so this is hopefully not a problem. If somebody can still reproduce this problem, please reopen and provide the information as said in comment #4.