On mips64, the compiler does not allow use of non-zero argument with
__builtin_frame_address(). However, the returned frame address is only
used when PIPE_ARCH_X86 is defined. The compile error can be avoided
by making #ifdef PIPE_ARCH_X86 cover the getting of frame address too.
The argument checking of __builtin_frame_address() has been present
as a debug assert in clang 8. In clang 10, there is a proper runtime
check for the argument. This is why the build has not failed before.
OK jsg@
instructions. Clang doesn't allow redeclaration (and therefore
redefinition) of the __sync_* builtins. Use #pragma redefine_extname
to work around that restriction. Clang also turns __sync_add_and_fetch
into __sync_fetch_and_add (and __sync_sub_and_fetch into
__sync_fetch_and_sub) in certain cases, so provide these functions as
well.
ok jsg@
When not in c++ or c11 mode:
- check for _Static_assert support in clang with __has_extension
- gcc implements _Static_assert starting with 4.6.0
- as a fallback, use a forward decl instead of ((void)0) so that
static_assert can be used at file scope (scope issue pointed out by
guenther@)
ok jsg@
autoconf function tests run without --std and have c11 visibility
timespec_get(3), a c11 function is found this way.
With gcc < 4.6 Mesa will build with --std=gnu99 with clang it will
build with --std=c99 and timespec_get() will no longer be visibile.
Further complicating things gcc 4.2 does not recognise --std=c11.
This was not noticed on linux as they build with _GNU_SOURCE which
among other things gives c11 visibility even with --std=c99.
discussed with millert@
issetugid(), not just the sloppy uid != euid test. gid != egid cases
can occur also.
Part of 6.6/009_mesaxlock.patch.sig and 6.5/020_mesaxlock.patch.sig
From Qualys, ok
From Brian Paul
02c3dad0f3b4d26e0faa5cc51d06bc50d693dcdc in mainline Mesa
"A security advisory (TALOS-2019-0857/CVE-2019-5068) found that
creating shared memory regions with permission mode 0777 could allow
any user to access that memory. Several Mesa drivers use shared-
memory XImages to implement back buffers for improved performance.
This path changes the shmget() calls to use 0600 (user r/w).
Tested with legacy Xlib driver and llvmpipe."
The intel driver can be used with inteldrm on ivy bridge or newer.
The radeon driver only works with amdgpu not radeondrm.
As we can't use python in xenocara add phony targets to create the
same output as python scripts which create json files for the loader.
This is just the ICDs, to use vulkan the loader and additional ports
are required.
ok mpi@ on an earlier version. Changed to call shm_unlink()
directly after shm_mkstemp() as suggested by kettenis@
shared libLLVM are in base. And now we can finally build the radeonsi
Mesa driver adding accelerated support for radeon parts based on the
Graphics Core Next (GCN) architecture which is everything since
GFX6 / Southern Islands.
This can later be expanded to other architectures which build libLLVM
and have clang as the default compiler (to handle LLVM's c++11 headers).
avoid a text relocation fixing the build with lld.
This time store the address of the GOT in ebx as required before calling
the PLT stub and change .balign value to match X86_ENTRY_SIZE as suggested
by naddy.
ok naddy@
Mesa 18.x needs an ld with build-id for at least the intel code
Mesa 18.2 assumes linux only memfd syscalls in intel code
Tested by matthieu@, kettenis@ and myself on a variety of hardware and
architectures. ok kettenis@
