Should we be concerned about this?...
src/platforms/android/server/buffer.cpp:
//TODO: we should make use of the android egl fence extension here to
update the fence.
// if the extension is not available, we should pass out a token that
the user
// will have to keep until
OK, I think this might be the problem. The client API is releasing its
shared pointer the the software buffer region prematurely (before
mir_buffer_stream_get_graphics_region even returns). This means we are
calling Android's gralloc->unlock() way before the client has even
started filling the
Hmm, no. BufferStream::secured_region seems to hold it for us.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1406725
Title:
Severe graphical corruption running software clients (including Xmir)
I'm fairly confident the problem is entirely in Mir. Not in Xmir.
Especially given it's easy to reproduce with mir-demos.
** Changed in: xorg-server (Ubuntu)
Status: Triaged => Invalid
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to
@Daniel,
Yeah, those flags are used by the implementations to know when they need
to do cache coherency operations (flush, invalidate, etc).
In android, software buffers CAN be used as backing to textures for
opengl and can be passed directly to Hardware Composer HAL (HW
Composer), hence they
Is it safe to mix HW and SW flags?
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1406725
Title:
Severe graphical corruption running software clients (including Xmir)
on android
To manage
Alberto's previous fix is here:
int mga::AndroidAllocAdaptor::convert_to_android_usage(BufferUsage usage)
{
switch (usage)
{
case mga::BufferUsage::use_hardware:
return (GRALLOC_USAGE_HW_TEXTURE | GRALLOC_USAGE_HW_RENDER);
case mga::BufferUsage::use_framebuffer_gles:
I recall Alberto fixed a corruption bug just like this using magic HWC
flags on the affected buffer. Not sure where that change went or if we
still have it.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
