[Mesa-dev] [PATCH] Fix Android 5.1 build and runtime issues

[Chad Versace]

On Tue 28 Jul 2015, Emil Velikov wrote:
> Hello Jim,
> 
> On 28 July 2015 at 02:57, Bish, Jim <jim.bish at intel.com> wrote:
> > From: Jim Bish <jim.bish at intel.com>
> >
> Would you mind splitting this into separate patches ? Adding a few
> words in the commit log(s) would be highly preferable.

Yes, please submit separate patches. "One patch does one thing, and does
that one thing well".

> 
> > ---
> >  Android.common.mk                           | 10 ++++++++++
> >  Android.mk                                  |  3 +++
> >  src/mesa/Android.libmesa_dricore.mk         |  2 +-
> >  src/mesa/drivers/dri/i965/intel_fbo.c       |  6 +++++-
> >  src/mesa/drivers/dri/i965/intel_tex_image.c |  4 +++-
> >  5 files changed, 22 insertions(+), 3 deletions(-)
> >
> > diff --git a/Android.common.mk b/Android.common.mk
> > index d662d60..68c2e1b 100644
> > --- a/Android.common.mk
> > +++ b/Android.common.mk
> > @@ -76,6 +76,16 @@ LOCAL_CFLAGS += \
> >         -D__STDC_LIMIT_MACROS
> >  endif
> >
> > +# add libdrm if there are hardware drivers
> > +ifneq ($(filter-out swrast,$(MESA_GPU_DRIVERS)),)
> > +LOCAL_CFLAGS += -DHAVE_LIBDRM
> 
> > +LOCAL_CFLAGS += \
> > +        -I$(DRM_TOP)/include/drm \
> > +        -I$(DRM_TOP) \
> > +        -I$(DRM_GRALLOC_TOP)

> The above includes should _not_ be needed. Please make sure that
> LOCAL_EXPORT_C_INCLUDE_DIRS for libdrm and friends are set correctly.
> The former two are addressed upstream [1] while there is a patch for
> the last one on mesa-dev (as there is no upstream for drm_gralloc) [2]
> 
> Do you know of any plans to use upstream libdrm ? Or at least
> rebase/merge with upstream changes ?
> 
> 
> > +LOCAL_SHARED_LIBRARIES += libdrm
> > +endif
> > +

I'm not an Android.mk expert, but this feels too high-level to add
libdrm to LOCAL_SHARED_LIBRARIES. I suspect this should be done in the
i965 directory instead.

> The rest of the hunk seems ok, but please mention why in the commit log.
> 
> >  LOCAL_CPPFLAGS += \
> >         $(if $(filter true,$(MESA_LOLLIPOP_BUILD)),-D_USING_LIBCXX) \
> >         -Wno-error=non-virtual-dtor \
> > diff --git a/Android.mk b/Android.mk
> > index ed160fb..8f523bb 100644
> > --- a/Android.mk
> > +++ b/Android.mk
> > @@ -45,6 +45,9 @@ endif
> >  MESA_COMMON_MK := $(MESA_TOP)/Android.common.mk
> >  MESA_PYTHON2 := python
> >
> > +DRM_TOP := external/drm
> > +DRM_GRALLOC_TOP := hardware/drm_gralloc
> > +
> Please don't. See above for details.
> 
> >  classic_drivers := i915 i965
> >  gallium_drivers := swrast freedreno i915g ilo nouveau r300g r600g radeonsi vmwgfx vc4
> >
> > diff --git a/src/mesa/Android.libmesa_dricore.mk b/src/mesa/Android.libmesa_dricore.mk
> > index 2e308b8..fef76c8 100644
> > --- a/src/mesa/Android.libmesa_dricore.mk
> > +++ b/src/mesa/Android.libmesa_dricore.mk
> > @@ -50,7 +50,7 @@ endif # MESA_ENABLE_ASM
> >  ifeq ($(ARCH_X86_HAVE_SSE4_1),true)
> >  LOCAL_SRC_FILES += \
> >         main/streaming-load-memcpy.c \
> > -       mesa/main/sse_minmax.c
> > +       main/sse_minmax.c
> >  LOCAL_CFLAGS := \
> >         -msse4.1 \
> >         -DUSE_SSE41
> > diff --git a/src/mesa/drivers/dri/i965/intel_fbo.c b/src/mesa/drivers/dri/i965/intel_fbo.c
> > index 87ba624..e6cdd76 100644
> > --- a/src/mesa/drivers/dri/i965/intel_fbo.c
> > +++ b/src/mesa/drivers/dri/i965/intel_fbo.c
> > @@ -355,13 +355,17 @@ intel_image_target_renderbuffer_storage(struct gl_context *ctx,
> >                                               screen->loaderPrivate);
> >     if (image == NULL)
> >        return;
> > -
> > +#ifndef HAVE_ANDROID_PLATFORM
> > +   /*
> > +    * Disable this check for Android
> > +    */
> >     if (image->planar_format && image->planar_format->nplanes > 1) {
> >        _mesa_error(ctx, GL_INVALID_OPERATION,
> >              "glEGLImageTargetRenderbufferStorage(planar buffers are not "
> >                 "supported as render targets.");
> >        return;
> >     }
> > +#endif
> Imho this should be a separate patch, with the commit and/or comment
> providing more information.

This should be in a separate patch. But...

You can't disable this check! Not without a heavy justification. If an
application binds a planar EGLImage as renderbuffer storage, does i965
correctly translate and scatter the fragment shader's output color to
the separate planes?  If not, what *does* i965 do?

Add some comments explaining that issue in the patch.

> 
> >
> >     /* __DRIimage is opaque to the core so it has to be checked here */
> >     switch (image->format) {
> > diff --git a/src/mesa/drivers/dri/i965/intel_tex_image.c b/src/mesa/drivers/dri/i965/intel_tex_image.c
> > index 3611280..4b3e9cf 100644
> > --- a/src/mesa/drivers/dri/i965/intel_tex_image.c
> > +++ b/src/mesa/drivers/dri/i965/intel_tex_image.c
> > @@ -317,8 +317,10 @@ intel_image_target_texture_2d(struct gl_context *ctx, GLenum target,
> >     if (image == NULL)
> >        return;
> >
> > +#ifndef HAVE_ANDROID_PLATFORM
> >     /* We support external textures only for EGLImages created with
> >      * EGL_EXT_image_dma_buf_import. We may lift that restriction in the future.
> > +    * We have to remove this restriction for Android builds.
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Why? That deserves an explanation.

> >      */
> >     if (target == GL_TEXTURE_EXTERNAL_OES && !image->dma_buf_imported) {
> >        _mesa_error(ctx, GL_INVALID_OPERATION,
> > @@ -326,7 +328,7 @@ intel_image_target_texture_2d(struct gl_context *ctx, GLenum target,
> >                 "for images created with EGL_EXT_image_dma_buf_import");
> >        return;
> >     }
> > -
> > +#endif

> Ditto.
> 
> Chad,
> I was under the impression that both restrictions were lifted with
> your EGL_EXT_image_dma_buf_import
> /OES_EGL_image_external work. Would you know, off hand, what is still missing ?

I didn't lift all restrictions. I lifted the restriction that
prevented glEGLImageTargetRenderbufferStorage and
glEGLImageTargetTexture2D from using non-planar dma_buf-backed EGLImages
as storage. In other words, you can now import an RGB EGLImage, render
to it, and texture from it. See commit a76dc15b2b3.

I did not alter restrictions related to planar formats.