Debian Patches

Status for yuzu/0-1734+ds-9

Patch Description Author Forwarded Bugs Origin Last update
cmake-mbedtls.patch Make CMake use system MbedTLS 3.6

Links to system MbedTLS and ports code to version 3		 Andrea Pappacoda <tachi@debian.org> no 2025-01-11
dynarmic-6.7.patch Port to dynarmic 6.7

Dynarmic 6.7 has introduced a breaking change by changing
UserConfig::fastmem_pointer from a pointer to an
std::optional<std::uintptr_t>.		 Andrea Pappacoda <andrea@pappacoda.it> no 2025-01-13
gamemode-linux-only.patch Only enable gamemode on Linux

Without this patch, gamemode is enabled on all UNIX-like systems like
BSDs and Hurd, which is of course wrong as gamemode is Linux-only.		 Andrea Pappacoda <andrea@pappacoda.it> no 2025-02-11
disable-mcl-library.patfch Disable linking with the mcl library

It doesn't seem to be used anywhere.		 Andrea Pappacoda <andrea@pappacoda.it> no 2025-02-11
boost-1.88.patch Fix building against Boost 1.88.0 Andrea Pappacoda <tachi@debian.org> not-needed debian backport, https://codeberg.org/yuzu-emu/yuzu/commit/e2da9359171daf5e3b3fdcb36ba7f34200f3e74d 2025-08-11
debian-git-revision.patch Disable automatic git revision

The values are defined in debian/rules		 Andrea Pappacoda <andrea@pappacoda.it> not-needed 2022-12-20
cmake-llvm-any-version.patch Accept any LLVM version

Otherwise, LLVM 19 isn't accepted, even if compatible		 Andrea Pappacoda <andrea@pappacoda.it> no 2025-01-11
0008-test-skip-RingBuffer-Threaded-Test.patch test: skip ~RingBuffer: Threaded Test
Broken with catch2 versions without thread safe assertions, added in
catch2 3.9.0:
https://github.com/catchorg/Catch2/commit/2a8a8a7210513fc2f6ae5e65986c0f2d730c1537		 Andrea Pappacoda <tachi@debian.org> no 2025-09-20
0009-arm64-fix-integer-underflow-in-ChooseVirtualBase.patch arm64: fix integer underflow in ChooseVirtualBase
When virtual_size >= Map39BitSize (1 << 39), the expression
(Map39BitSize - virtual_size) wraps around as an unsigned subtraction,
producing a huge value for `upper`. The subsequent `range = upper -
lower` also wraps, so all 64 hint addresses are garbage; mmap() never
returns one at the requested hint, and ChooseVirtualBase() returns
MAP_FAILED every time.

The HostMemory outer constructor catches the resulting std::bad_alloc
and falls back to a plain VirtualBuffer, setting virtual_base = nullptr.
Map() then silently returns early (it guards on !virtual_base), so the
virtual-memory pointer returned by VirtualBasePointer() is null. The
test then crashes writing to nullptr + 0x5000 = 0x5000.

The test uses VIRTUAL_SIZE = 1ULL << 39 = Map39BitSize. The Impl
constructor pads this by one HugePageSize (to allow later HugePageSize
alignment of virtual_base), so the actual size passed to
ChooseVirtualBase is always > Map39BitSize.

virtual_size < Map39BitSize *and* the resulting [lower, upper) range is
non-empty (upper > lower). Fall back to an unconstrained mmap(nullptr)
otherwise, which succeeds on the 48-bit VA arm64 Debian build hosts.		 Bastian Germann <bage@debian.org> no 2026-05-02
0010-build-also-ask-for-GuiPrivate-Qt-component.patch build: also ask for GuiPrivate Qt component
Qt 6.10.0 now no longer automatically makes available the private
headers of a given component, but requires to explicitly ask for the
Private counterpart. Introduced in upstream commit ad7b94e163a:
<https://code.qt.io/cgit/qt/qtbase.git/commit/?id=ad7b94e163ac5c3959a7e38d7f48536be288a187>		 Andrea Pappacoda <tachi@debian.org> no 2026-05-03

All known versions for source package 'yuzu'

Links