http://bugs.winehq.org/show_bug.cgi?id=29194
--- Comment #7 from Daniel Calviño Sánchez danxuliu@gmail.com 2011-12-20 18:33:49 CST --- Created attachment 38055 --> http://bugs.winehq.org/attachment.cgi?id=38055 Differences between traces from Catalyst 11.5 (working) and Catalyst 11.12 (not working)
I have found something that perhaps can give you a hint about the problem. I have executed the demo with Catalyst 11.5 (the last release where it worked) and Catalyst 11.12 (the latest one) with the same version of Wine (1.3.35, the latest one) and compared the trace. There are only three* differences between them.
The trace with Catalyst 11.5 was trimmed down manually to the point where the trace with Catalyst 11.12 began to repeat the same output over and over. Then, I replaced all the hexadecimal values in the traces by the word "hexadecimal", as they are mainly memory addresses and thus different in each run, using sed (sed -r "s/0x[0-9a-f]+/hexadecimal/g"). Finally, I got the difference between the Catalyst 11.5 trace and the Catalyst 11.12 trace with diff command.
When Catalyst 11.12 was used, the trace showed two errors that did not appear in the Catalyst 11.5 trace: "err:d3d:context_create wglSwapIntervalEXT failed to set swap interval 0 for context 0x1d1d80, last error 0x591" and "err:d3d:context_create wglSwapIntervalEXT failed to set swap interval 0 for context 0x1c6524b8, last error 0x591". The context value in this second error is the same that appears in the lines repeated over and over.
The other difference in the trace is that, even before the point where Catalyst 11.12 repeats the same output over and over, Catalyst 11.5 has 1392 lines that do not appear in the Catalyst 11.12 trace.
I have attached the diff between both traces and a context of 105 lines. I have used this huge context to reach the "end" (where Catalyst 11.12 starts to output the same 8 lines over and over) of the trace after the bunch of Catalyst 11.5 only lines.
*Well, there is another difference, which is a different thread number in ony point, but I think that it changes between executions, like memory addresses.