On 20 November 2016 at 16:28, Ilia Mirkin <imir...@alum.mit.edu> wrote: > Thanks Marek. > > Emil, I'd appreciate it if you could have a look as well. I believe > you were the last to touch these helpers. I'm highly confused by the > flow of things, and mainly want to ensure that we don't warp the same > screen *twice*. > Thanks for the poke.
>>> diff --git a/src/gallium/auxiliary/target-helpers/inline_sw_helper.h >>> b/src/gallium/auxiliary/target-helpers/inline_sw_helper.h >>> index 5bb77a5..3d9d824 100644 >>> --- a/src/gallium/auxiliary/target-helpers/inline_sw_helper.h >>> +++ b/src/gallium/auxiliary/target-helpers/inline_sw_helper.h >>> @@ -55,6 +55,9 @@ sw_screen_create_named(struct sw_winsys *winsys, const >>> char *driver) >>> screen = swr_create_screen(winsys); >>> #endif >>> >>> + if (screen) >>> + screen = debug_screen_wrap(screen); >>> + Hmm things are pretty nice in here. We cannot have this hunk since every user (old school static only modules and pipe-foo) explicitly wraps the screen. >>> return screen; >>> } >>> >>> diff --git a/src/gallium/auxiliary/target-helpers/sw_helper.h >>> b/src/gallium/auxiliary/target-helpers/sw_helper.h >>> index 5e4e9f7..d1945ad 100644 >>> --- a/src/gallium/auxiliary/target-helpers/sw_helper.h >>> +++ b/src/gallium/auxiliary/target-helpers/sw_helper.h >>> @@ -57,6 +57,9 @@ sw_screen_create_named(struct sw_winsys *winsys, const >>> char *driver) >>> screen = swr_create_screen(winsys); >>> #endif >>> >>> + if (screen) >>> + screen = debug_screen_wrap(screen); >>> + Option A: is to keep only the sw_helper.h change. Option B: --- a/src/gallium/auxiliary/pipe-loader/pipe_loader_sw.c +++ b/src/gallium/auxiliary/pipe-loader/pipe_loader_sw.c @@ -301,9 +301,16 @@ pipe_loader_sw_create_screen(struct pipe_loader_device *dev) screen = sdev->dd->create_screen(sdev->ws); if (!screen) - sdev->ws->destroy(sdev->ws); + goto fail; +#if GALLIUM_STATIC_TARGETS + screen = debug_screen_wrap(screen); +#endif return screen; + +fail: + sdev->ws->destroy(sdev->ws); + return NULL; } If you like A, please add a comment - feel free to reuse mine. I'm fine with either solution, pre-emptively Reviewed-by: Emil Velikov <emil.veli...@collabora.com> -Emil P.S. In general the wrapper drivers are pretty inconsistent (nasty?) - some error out, others continue, iirc some could even crash in OOM conditions. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
