| Summary: | Only black content on screen, in the Tokyo flashback of the game "The Secret World" | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | John <john.ettedgui> |
| Component: | Drivers/Gallium/radeonsi | Assignee: | Default DRI bug account <dri-devel> |
| Status: | CLOSED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | medium | ||
| Version: | git | ||
| Hardware: | x86-64 (AMD64) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
|
Description
John
2014-09-11 06:37:36 UTC
If you could create an apitrace demonstrating the problem, that should be useful. Thank you for the prompt reply Michel. This is my first time with apitrace so I may have made a mistake. Here's the command line I've used: apitrace32 trace --api=gl /usr/bin/wine "<game>.exe". I've tried running apitrace in 64bit but it never worked, I suppose because the game is 32bit. The result files are huge, is this expected? 5.3G, 327k, 236M I can't see how I could share those with you. (well apart from the 327k one). Here is a trim frame I got running this command: apitrace trim --exact --frames=1236 wine-preloader.1.trace and then compressed. The frame 1236 is the in-game blacked out case, seen in qapitrace. https://mega.co.nz/#!s4RAAAhL!kjBFx0Yq-QgvQTtflnPoRsbOQpdYTBAwWDBTrlHuSLY I've done the same with frame 1183, which I believe is in the cinematic but since that one is completely black it's hard to really say for sure. https://mega.co.nz/#!49RySQhJ!jh6t2rDLrKDzcSNUKViWBwrpn10vNcuPGMu8QuDbXCw I've used mega as the files are too big to attach to the bug report. Thanks! I've tried using the nine state tracker and it doesn't fix this issue (but it does run the game quite faster) I'm afraid these traces are trimmed down too much, I can't replay them: 0 @2 glBindBufferARB(target = GL_ARRAY_BUFFER, buffer = 17) 0: warning: no current context 1 @2 glMapBufferRange(target = GL_ELEMENT_ARRAY_BUFFER, offset = 0, length = 20480, access = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT) = 0xee18d000 1: warning: no current context glretrace: ../retrace/retrace_swizzle.cpp:137: void retrace::addRegion(long long unsigned int, void*, long long unsigned int): Assertion `buffer' failed. How would I trim them less ? Did I use a wrong command? Thanks! I don't have much experience with trimming apitraces, so I'm afraid I can't help you with that. That said, if you can include the first frame in the trimmed apitrace, that might help. However, this will leave some uncertainty about whether any issues demonstrated by the other frame aren't caused by the apitrace trimming working incorrectly... Well, I tried something else, I probably should have tried earlier: Compressing the original trace with xz at its maximum compression level, and it is very acceptable. (more than 5 Gb before vs less than 200Mb after). The link: https://mega.co.nz/#!khgwBJRC!Xzk1fsNeKZgUS5aqdQ3cqDWkGQoynGj3TPYyr0XGdGE Locating the issues in the trace: 1- After I load my character, you can see for just a little bit some sort of Zombie image with some text, then it switches to all black. I do not really care about that all back, though it probably means something as well. This is just the loading phase going all black. What I care about is right after, it is still all black, but now you can see the FPS counter of the left side. That is the in-game cinematic all black. If you pay careful attention, and have proper lighting in your room, you may be able to see characters moving and such (all black of course, you can only see them as shadows) 2- After the cinematic, you'll see a game UI but everything else is all black, that is the second scene with issue. Please do not worry about the original weird stuff happening in the intro, when the trace begins, it is part of the game. Thank you! Just in case, I tried with Alex's linux drm-next tree, but it didn't make a difference. I was hoping Marek's latest big patchset would help, but it didn't... (well maybe the game was a little faster but that is not what I am looking for here). I tried with latest git, but still the same. This seems to work correctly with current Git of Mesa and LLVM. I have just tested it and you are correct, it works with latest git. The performance and texture quality are not there yet, but that's a different issue (that could just be a wine one anyway...). Thank you for fixing this, and remembering the bug! How did you know it was fixed? (In reply to John from comment #14) > How did you know it was fixed? I didn't know, I just guessed it might be fixed thanks to some fixes going in recently. Well, good guess! |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.