Message ID | 20191015213101.27460-4-lothar.felten@gmail.com |
---|---|
State | Superseded |
Headers | show |
Series | [RFC,1/4] new package: ti-sgx/ti-sgx-libgbm libgbm for SGX graphics accelerator | expand |
On 15/10/2019 23:31, Lothar Felten wrote: > qt5 configure options: add ti-sgx-libgbm as GBM backend > > Signed-off-by: Lothar Felten <lothar.felten@gmail.com> > --- > package/qt5/qt5base/qt5base.mk | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/package/qt5/qt5base/qt5base.mk b/package/qt5/qt5base/qt5base.mk > index e4aa76908..0caa57000 100644 > --- a/package/qt5/qt5base/qt5base.mk > +++ b/package/qt5/qt5base/qt5base.mk > @@ -74,10 +74,13 @@ else > QT5BASE_CONFIGURE_OPTS += -no-kms > endif > > -# Uses libgbm from mesa3d > +# Uses libgbm from mesa3d or ti-sgx-libgbm > ifeq ($(BR2_PACKAGE_MESA3D_OPENGL_EGL),y) > QT5BASE_CONFIGURE_OPTS += -gbm > QT5BASE_DEPENDENCIES += mesa3d > +else ifeq ($(BR2_PACKAGE_TI_SGX_LIBGBM),y) I would swap the two so TI gets priority if both are selected. Note that there was also talk about making gbm a virtual package (with currently mesa3d and ti-sgx-libgbm as providers). IIRC, though, the TI gbm package was not 100% compatible. But apparently it's good enough to build qt5 then? And what about waffle and mesa3d-demos? Note that we don't require you to figure all of this out before we can apply these patches. But maybe you know a little bit more about it. Regards, Arnout > +QT5BASE_CONFIGURE_OPTS += -gbm > +QT5BASE_DEPENDENCIES += ti-sgx-libgbm > else > QT5BASE_CONFIGURE_OPTS += -no-gbm > endif >
Hello, On Wed, 16 Oct 2019 01:00:37 +0200 Arnout Vandecappelle <arnout@mind.be> wrote: > Note that there was also talk about making gbm a virtual package (with > currently mesa3d and ti-sgx-libgbm as providers). IIRC, though, the TI gbm > package was not 100% compatible. But apparently it's good enough to build qt5 then? I tried to make libgbm a virtual package, but unfortunately my conclusion was that the libgbm API is not really standardized. When I tried to build kmscube against the TI gbm, it would fail miserably, and it was also failing with another non-mesa GBM implementation. I pushed what I had back then at https://github.com/tpetazzoni/buildroot/commits/gbm. > Note that we don't require you to figure all of this out before we can apply > these patches. But maybe you know a little bit more about it. Yes, let's not make this depend on having a virtual package for libgbm. We can figure that out later. Thomas
diff --git a/package/qt5/qt5base/qt5base.mk b/package/qt5/qt5base/qt5base.mk index e4aa76908..0caa57000 100644 --- a/package/qt5/qt5base/qt5base.mk +++ b/package/qt5/qt5base/qt5base.mk @@ -74,10 +74,13 @@ else QT5BASE_CONFIGURE_OPTS += -no-kms endif -# Uses libgbm from mesa3d +# Uses libgbm from mesa3d or ti-sgx-libgbm ifeq ($(BR2_PACKAGE_MESA3D_OPENGL_EGL),y) QT5BASE_CONFIGURE_OPTS += -gbm QT5BASE_DEPENDENCIES += mesa3d +else ifeq ($(BR2_PACKAGE_TI_SGX_LIBGBM),y) +QT5BASE_CONFIGURE_OPTS += -gbm +QT5BASE_DEPENDENCIES += ti-sgx-libgbm else QT5BASE_CONFIGURE_OPTS += -no-gbm endif
qt5 configure options: add ti-sgx-libgbm as GBM backend Signed-off-by: Lothar Felten <lothar.felten@gmail.com> --- package/qt5/qt5base/qt5base.mk | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)