Use the previously discussed soc_id which have public values in EDK2.
The so far used toplevel devicetree compatible values weren't as precise
and some logic (x1, x2p, x2e) simply was dead code.
Also drops the CPU test due to reading soc_id sysfs file, could be
adapted if still wanted around.
* Vulkan: fixes driver name concatenation on multi-GPU systems
applyDriverName appended to the result buffer without clearing it. When
a later device reported a higher API version, its driver name was
appended to the name of the earlier device instead of replacing it, as
the comment in detectVulkan intends.
* CHANGELOG: moves the #2615 entry to Unreleased
#2616 was opened after 2.69.0 had been released, but its entry was
written under 2.69.0 Bugfixes, so the changelog listed the fix as part
of 2.69.0.
DDC/CI results never set `builtin`, and FF_LIST_ADD does not zero the new
element, so external monitors were reported as built-in or external at
random. Set it to false, as the BSD DDC/CI backend does.
FF_SUPPRESS_IO() was declared inside the `if (ffddca_init)` block, so its
cleanup restored stdout before display detection ran and libddcutil trace
lines could reach the output, ahead of `--format json` too. Declare it at
function scope so it covers the whole detection, including the
FF_DISABLE_DLOPEN build.
Closes#2615
FreeBSD 16.0-CURRENT's <string.h> now provides C23-style
qualifier-preserving wrappers for the string search functions. Besides
the prototype, `memrchr` (like `memchr`, `strchr`, `strstr`, `memmem`,
...) is also defined as a function-like macro built on `_Generic`:
#define memrchr(b, c, n) __qualsel((b), \
(const void *)(memrchr)((b), (c), (n)), (memrchr)((b), (c), (n)))
`common/memrchr.h` redeclares the function unconditionally. When
<string.h> has already been included, that declaration is expanded by
the macro and every translation unit fails with:
src/common/memrchr.h:11:7: error: expected identifier or '('
11 | void* memrchr(const void* s, int c, size_t n);
A libc that defines `memrchr` as a macro has necessarily declared the
function it wraps, so skip our declaration in that case. Nothing
changes on platforms where `memrchr` is not a macro, and the fallback
implementation is still selected by `check_function_exists`.
Carter Li