Re: Одговор: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Am 06.06.2009 um 08:18 schrieb Roderick Colenbrander: We don't use GL extensions when they aren't around. Perhaps there is a small check which fails for FBOs. Backbuffer should work fine but a lot more changes have been made in .23 and those are which cause the zbuffer error. A long time ago there was code that fell back to Pbuffers if FBOs were enabled, but not supported - that PBuffer fallback was used even if pbuffers were not available. I think I fixed that though
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
2009/6/6 Stefan Dösinger : > Hmm. Does fglrx support this extension at all? http://bugs.winehq.org/attachment.cgi?id=21581 That's attached to bug 18794.
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
That's probably the same issue as http://bugs.winehq.org/show_bug.cgi?id=18794 I'm afraid the conclusion there is simply going to be that fglrx's FBO support sucks. We could probably justify not checking depth stencil formats, but it's perfecly reasonable to try GL_COMPRESSED_RED_GREEN_RGTC2_EXT. The driver should of course crash on neither of those. Hmm. Does fglrx support this extension at all? I think fglrx has GL_ATI_texture_compression_3dc, which is essentially the same, but uses GL_COMPRESSED_LUMINANCE_ALPHA_3DC_ATI. Besides that, I don't think I have ever seen renderable compressed textures anywhere, so I'd say it is reasonable not to test it either. Still, the driver shouldn't crash.
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
2009/6/6 Jerome Leclanche : > Would it be a good/stupid idea to check for fglrx during wineboot, and > set OSRM to a different value than fbo? > That's essentially what blacklisting EXT_framebuffer_object would do, although during wined3d initialization, not wineboot. We want to avoid that for the reasons already mentioned by Roderick. Not using FBOs (EXT_framebuffer_blit in particular) also has a significant performance impact in a number of applications. HL2 and CS:S are good examples.
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Sure that is possible but it is something we should avoid. In the longterm having this enabled by default is better as it would get ATI and others to fix their drivers. If we would restore the old value (the old method is also not great for modern games and causes a lot of issues, FBOs are needed for proper d3d9 support) it would take longer for ATI to fix the remaining driver bugs. Roderick On Sat, Jun 6, 2009 at 12:32 PM, Jerome Leclanche wrote: > Would it be a good/stupid idea to check for fglrx during wineboot, and > set OSRM to a different value than fbo? > > 2009/6/6 Henri Verbeet : >> 2009/6/6 Kovács András : >>> wine: Unhandled page fault on read access to 0x0018 at address >>> 0x7c71fb02 (thread 0009), starting debugger... >>> Unhandled exception: page fault on read access to 0x0018 in 32-bit >>> code (0x7c71fb02). >>> Register dump: >>> CS:0073 SS:007b DS:007b ES:007b FS:0033 GS:003b >>> EIP:7c71fb02 ESP:0033e2a4 EBP:7d838810 EFLAGS:00210246( R- -- I Z- >>> -P- ) >>> EAX:0001 EBX:0001 ECX:7d49d6a0 EDX: >>> ESI: EDI:7d715258 >>> Stack dump: >>> 0x0033e2a4: 7c70ca7f 7d847520 >>> 0x0033e2b4: 7d612f90 7d612f90 >>> 0x0033e2c4: 7d82e7b8 7d612f90 >>> 0x0033e2d4: 7d82e7b8 7d848010 7d82e7b8 >>> 0x0033e2e4: 7d7e2628 >>> 0x0033e2f4: 7d612f90 7d6a7e88 >>> Backtrace: >>> =>0 0x7c71fb02 in fglrx_dri.so (+0x713b02) (0x7d838810) >>> 0x7c71fb02: movl 0x18(%edx),%eax >>> >> >> That's probably the same issue as >> http://bugs.winehq.org/show_bug.cgi?id=18794 >> I'm afraid the conclusion there is simply going to be that fglrx's FBO >> support sucks. We could probably justify not checking depth stencil >> formats, but it's perfecly reasonable to try >> GL_COMPRESSED_RED_GREEN_RGTC2_EXT. The driver should of course crash >> on neither of those. >> >> I suppose we could blacklist EXT_framebuffer_object on fglrx, but that >> will likely just hurt more in the long term. >> >> >> > > > > -- > J. Leclanche / Adys > > >
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Would it be a good/stupid idea to check for fglrx during wineboot, and set OSRM to a different value than fbo? 2009/6/6 Henri Verbeet : > 2009/6/6 Kovács András : >> wine: Unhandled page fault on read access to 0x0018 at address >> 0x7c71fb02 (thread 0009), starting debugger... >> Unhandled exception: page fault on read access to 0x0018 in 32-bit >> code (0x7c71fb02). >> Register dump: >> CS:0073 SS:007b DS:007b ES:007b FS:0033 GS:003b >> EIP:7c71fb02 ESP:0033e2a4 EBP:7d838810 EFLAGS:00210246( R- -- I Z- >> -P- ) >> EAX:0001 EBX:0001 ECX:7d49d6a0 EDX: >> ESI: EDI:7d715258 >> Stack dump: >> 0x0033e2a4: 7c70ca7f 7d847520 >> 0x0033e2b4: 7d612f90 7d612f90 >> 0x0033e2c4: 7d82e7b8 7d612f90 >> 0x0033e2d4: 7d82e7b8 7d848010 7d82e7b8 >> 0x0033e2e4: 7d7e2628 >> 0x0033e2f4: 7d612f90 7d6a7e88 >> Backtrace: >> =>0 0x7c71fb02 in fglrx_dri.so (+0x713b02) (0x7d838810) >> 0x7c71fb02: movl 0x18(%edx),%eax >> > > That's probably the same issue as http://bugs.winehq.org/show_bug.cgi?id=18794 > I'm afraid the conclusion there is simply going to be that fglrx's FBO > support sucks. We could probably justify not checking depth stencil > formats, but it's perfecly reasonable to try > GL_COMPRESSED_RED_GREEN_RGTC2_EXT. The driver should of course crash > on neither of those. > > I suppose we could blacklist EXT_framebuffer_object on fglrx, but that > will likely just hurt more in the long term. > > > -- J. Leclanche / Adys
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
"chris ahrendt" wrote: Using this attached Reg File Is still a no go... same... (I forgot the Direct3D tag but added it and no change) Have you heard of a regression testing? Also wine-devel is not an appropriate place to report bugs. -- Dmitry.
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Using this attached Reg File Is still a no go... same... (I forgot the Direct3D tag but added it and no change) Chris rendering3.Reg Description: Binary data
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
2009/6/6 Kovács András : > wine: Unhandled page fault on read access to 0x0018 at address > 0x7c71fb02 (thread 0009), starting debugger... > Unhandled exception: page fault on read access to 0x0018 in 32-bit > code (0x7c71fb02). > Register dump: > CS:0073 SS:007b DS:007b ES:007b FS:0033 GS:003b > EIP:7c71fb02 ESP:0033e2a4 EBP:7d838810 EFLAGS:00210246( R- -- I Z- > -P- ) > EAX:0001 EBX:0001 ECX:7d49d6a0 EDX: > ESI: EDI:7d715258 > Stack dump: > 0x0033e2a4: 7c70ca7f 7d847520 > 0x0033e2b4: 7d612f90 7d612f90 > 0x0033e2c4: 7d82e7b8 7d612f90 > 0x0033e2d4: 7d82e7b8 7d848010 7d82e7b8 > 0x0033e2e4: 7d7e2628 > 0x0033e2f4: 7d612f90 7d6a7e88 > Backtrace: > =>0 0x7c71fb02 in fglrx_dri.so (+0x713b02) (0x7d838810) > 0x7c71fb02: movl 0x18(%edx),%eax > That's probably the same issue as http://bugs.winehq.org/show_bug.cgi?id=18794 I'm afraid the conclusion there is simply going to be that fglrx's FBO support sucks. We could probably justify not checking depth stencil formats, but it's perfecly reasonable to try GL_COMPRESSED_RED_GREEN_RGTC2_EXT. The driver should of course crash on neither of those. I suppose we could blacklist EXT_framebuffer_object on fglrx, but that will likely just hurt more in the long term.
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On 06/05/2009 09:16 PM, Austin English wrote: > On Fri, Jun 5, 2009 at 8:03 PM, chris ahrendt wrote: > >> On 06/05/2009 08:15 PM, Austin English wrote: >> >>> On Fri, Jun 5, 2009 at 7:11 PM, chris ahrendt >>> wrote: >>> >>> Tried it several times.. and different iterations of the registry keys (you will find them attached) and both cases it fails the same way any of the d3d tests will fail or the machine locks up. >>> >> >> > Use that attached .reg file. > > > Which one I attached two and neither of them worked at all.. Chris
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On Fri, Jun 5, 2009 at 8:03 PM, chris ahrendt wrote: > > On 06/05/2009 08:15 PM, Austin English wrote: >> On Fri, Jun 5, 2009 at 7:11 PM, chris ahrendt wrote: >> >>> Tried it several times.. and different iterations of the registry keys >>> (you will find them attached) and both cases it fails the same way any >>> of the d3d tests will fail or the machine locks up. >>> >> > Use that attached .reg file. -- -Austin
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Hi! With new wine (1.1.23+ git) I'm getting wide range of failures of different games. System: Ubuntu Jaunty x86 ATI Radeon HD4870 512 MB fglrx 9.5 - The most significant crash (with several games): wine: Unhandled page fault on read access to 0x0018 at address 0x7c71fb02 (thread 0009), starting debugger... Unhandled exception: page fault on read access to 0x0018 in 32-bit code (0x7c71fb02). Register dump: CS:0073 SS:007b DS:007b ES:007b FS:0033 GS:003b EIP:7c71fb02 ESP:0033e2a4 EBP:7d838810 EFLAGS:00210246( R- -- I Z- -P- ) EAX:0001 EBX:0001 ECX:7d49d6a0 EDX: ESI: EDI:7d715258 Stack dump: 0x0033e2a4: 7c70ca7f 7d847520 0x0033e2b4: 7d612f90 7d612f90 0x0033e2c4: 7d82e7b8 7d612f90 0x0033e2d4: 7d82e7b8 7d848010 7d82e7b8 0x0033e2e4: 7d7e2628 0x0033e2f4: 7d612f90 7d6a7e88 Backtrace: =>0 0x7c71fb02 in fglrx_dri.so (+0x713b02) (0x7d838810) 0x7c71fb02: movl0x18(%edx),%eax Inside ati driver. - Sims 3 exits without any error message. (it worked before with fbo's, now exits. it can be started with backbuffer setting) 2009. 06. 5, péntek keltezéssel 18.03-kor chris ahrendt ezt írta: > On 06/05/2009 08:15 PM, Austin English wrote: > > On Fri, Jun 5, 2009 at 7:11 PM, chris ahrendt wrote: > > > >> Tried it several times.. and different iterations of the registry keys > >> (you will find them attached) and both cases it fails the same way any > >> of the d3d tests will fail or the machine locks up. > >> > > > > > > > > > > > >
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On 06/05/2009 08:15 PM, Austin English wrote: > On Fri, Jun 5, 2009 at 7:11 PM, chris ahrendt wrote: > >> Tried it several times.. and different iterations of the registry keys >> (you will find them attached) and both cases it fails the same way any >> of the d3d tests will fail or the machine locks up. >> >
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On Fri, Jun 5, 2009 at 7:11 PM, chris ahrendt wrote: > Tried it several times.. and different iterations of the registry keys > (you will find them attached) and both cases it fails the same way any > of the d3d tests will fail or the machine locks up. -- -Austin REGEDIT4 [HKEY_CURRENT_USER\Software\Wine\Direct3D] "OffscreenRenderingMode"="backbuffer"
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Tried it several times.. and different iterations of the registry keys (you will find them attached) and both cases it fails the same way any of the d3d tests will fail or the machine locks up. rendering.Reg Description: Binary data rendering2.Reg Description: Binary data
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On 06/05/2009 05:10 PM, Roderick Colenbrander wrote: > Check the useful registry keys page at wiki.winehq.org (it is > Direct3D/OffscreenRenderingMethod) > > Roderick > > On Fri, Jun 5, 2009 at 10:41 PM, chris ahrendt wrote: > >> On 06/05/2009 04:32 PM, Roderick Colenbrander wrote: >> >>> As of Wine 1.1.23 wined3d made FBOs the default offscreen rendering >>> method. If you are using a non-nvidia card (those in general have >>> buggier drivers) that could explain all the failures. Setting >>> OffscreenRenderingMethod to backbuffer restores the old behavior. >>> >>> Roderick >>> >>> On Fri, Jun 5, 2009 at 10:04 PM, chris ahrendt >>> wrote: >>> >>> Something majorly changed and is wrong with the latest GIT tree. Between the comits yesterday and today 90% of the tests get exceptions or fail. All of the d3d tests get a process exception and fail with a dialog box. Ideas? Chris >>> >>> >> where do I set that? >> >> >> >> >> >> >> >> >> > > Added the registry key and no go... still failure... Chris
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
Check the useful registry keys page at wiki.winehq.org (it is Direct3D/OffscreenRenderingMethod) Roderick On Fri, Jun 5, 2009 at 10:41 PM, chris ahrendt wrote: > > On 06/05/2009 04:32 PM, Roderick Colenbrander wrote: >> As of Wine 1.1.23 wined3d made FBOs the default offscreen rendering >> method. If you are using a non-nvidia card (those in general have >> buggier drivers) that could explain all the failures. Setting >> OffscreenRenderingMethod to backbuffer restores the old behavior. >> >> Roderick >> >> On Fri, Jun 5, 2009 at 10:04 PM, chris ahrendt wrote: >> >>> Something majorly changed and is wrong with the latest GIT tree. >>> Between the comits yesterday and today 90% of the tests get exceptions >>> or fail. >>> All of the d3d tests get a process exception and fail with a dialog box. >>> >>> >>> Ideas? >>> >>> Chris >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >> >> > where do I set that? > > > > > > > >
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
On 06/05/2009 04:32 PM, Roderick Colenbrander wrote: > As of Wine 1.1.23 wined3d made FBOs the default offscreen rendering > method. If you are using a non-nvidia card (those in general have > buggier drivers) that could explain all the failures. Setting > OffscreenRenderingMethod to backbuffer restores the old behavior. > > Roderick > > On Fri, Jun 5, 2009 at 10:04 PM, chris ahrendt wrote: > >> Something majorly changed and is wrong with the latest GIT tree. >> Between the comits yesterday and today 90% of the tests get exceptions >> or fail. >> All of the d3d tests get a process exception and fail with a dialog box. >> >> >> Ideas? >> >> Chris >> >> >> >> >> >> >> >> >> >> >> > > where do I set that?
Re: Latest Git build 7aeffc442cc1 for Wine 1.1.23
As of Wine 1.1.23 wined3d made FBOs the default offscreen rendering method. If you are using a non-nvidia card (those in general have buggier drivers) that could explain all the failures. Setting OffscreenRenderingMethod to backbuffer restores the old behavior. Roderick On Fri, Jun 5, 2009 at 10:04 PM, chris ahrendt wrote: > > Something majorly changed and is wrong with the latest GIT tree. > Between the comits yesterday and today 90% of the tests get exceptions > or fail. > All of the d3d tests get a process exception and fail with a dialog box. > > > Ideas? > > Chris > > > > > > > > > >
Latest Git build 7aeffc442cc1 for Wine 1.1.23
Something majorly changed and is wrong with the latest GIT tree. Between the comits yesterday and today 90% of the tests get exceptions or fail. All of the d3d tests get a process exception and fail with a dialog box. Ideas? Chris