Repository navigation
[d3d10] fix d3d10core crashing every d3d10 game - #83
Conversation
d3d10core was calling D3D11CoreCreateDevice with the newer 10 argument version, but sarek's d3d11.dll still exports the old 6 argument one. The device ended up in the wrong spot, so every d3d10 game got a null device and crashed right away. Go back to the 6 argument call and link d3d10core against sarek's own d3d11 again.
yeah this was kinda my bad, it looked fine so i merged it. i don't have any d3d10 games to test with right now. i've got the whole range from d3d3 to d3d11 except d3d10, mostly because it was never that popular as an api to begin with. and since d3d10 in sarek is basically done and hasn't changed in a while, no point keeping d3d10 test games around if i don't play them and i dont need them for testing. gonna do a small release soon.
thanks for your help mate edit: you might want to use 1.14.0 or the new release instead of 1.13.0, as it fixes OOMs on several games on low memory systems |
Hey, every D3D10 game crashes on start with Sarek 1.13.0 and 1.14.0.
Since #41, d3d10core calls
D3D11CoreCreateDevicewith the newer 10 argument version, but Sarek's d3d11.dll still exports the old 6 argument one. So the device lands in the wrong spot, d3d10core gets back a null device and crashes on the next line.This puts the call back to the 6 argument version and links d3d10core against Sarek's own d3d11 again (the toolchain's libd3d11.a has the 10 argument one, which is why the 32-bit build needs this).
We've been shipping a patched 1.13.0 in Bannerlator and D3D10 games work again with it, everything else unchanged. If you'd rather update d3d11's export to the 10 argument version instead, that works too.
Thanks for keeping Sarek going!