Mesa 25.2.4 Release Notes / 2025-10-01

Mesa 25.2.4 is a bug fix release which fixes bugs found since the 25.2.3 release.

Mesa 25.2.4 implements the OpenGL 4.6 API, but the version reported by glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) / glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used. Some drivers don’t support all the features required in OpenGL 4.6. OpenGL 4.6 is only available if requested at context creation. Compatibility contexts may report a lower version depending on each driver.

Mesa 25.2.4 implements the Vulkan 1.4 API, but the version reported by the apiVersion property of the VkPhysicalDeviceProperties struct depends on the particular driver being used.

SHA checksums

SHA256: a370b4c549cbfbe646b319e34d73edb50ed883978f5e95133f282f0eae39ab52  mesa-25.2.4.tar.xz
SHA512: ec74e1cc101b68ad2462399554273a9debb3d93a05c0a315cde3259b250d5828cbcef7771049b52beb01d07c34122fa5257451409b1d2b7fdcbf055c0f2c863d  mesa-25.2.4.tar.xz

New features

  • None

Bug fixes

  • xe2: DMA Buffer exported modifier is incorrect

  • [ANV][LNL] - FINAL FANTASY XVI (2515020) - Title crashes to Desktop immediately following the splash card.

  • nvk, nak: NAK panic in Call of the Wild: The Angler on RTX 4080

  • Simple External Semaphore test hangs in vk_sync_wait

  • dEQP-VK.spirv_assembly.instruction.compute.float_controls.fp32.input_args.reflect_denorm_flush_to_zero regression on nvk

  • Crash on game Elite Dangerous at 0% planetary generation, on Tigerlake+ Iris Xe and Arc GPUs.

  • Gallium: Segfault while trying to compile a shader with differing UBO contents in fragment and vertex stage - With reproduction case

  • aco: generate wrong code when gl_DrawID is used by primitive indices in mesh shader

Changes

Bas Nieuwenhuizen (1):

  • device-select: Fix error check.

Calder Young (1):

  • anv: Fix tiling for AV1 IntraBC surface on Gfx125+

Christoph Neuhauser (1):

  • egl: Fix invalid device UUID returned by EGL_EXT_device_persistent_id

Daniel Schürmann (1):

  • aco/isel: rename emit_readfirstlane() -> emit_vector_as_uniform()

Danylo Piliaiev (1):

  • freedreno/decode: Fix preamble decoding

Dave Airlie (1):

  • gallivm: handle u8/u16 const loads properly on big-endian.

Ella Stanforth (1):

  • v3d/compiler: Lower load_output after logic operations

Eric Engestrom (4):

  • docs: add sha sum for 25.2.3

  • .pick_status.json: Update to 42a78a1aae19f855b049462d7714cd1f07ca12e4

  • .pick_status.json: Mark 59278c223614a3877fcbbabb8c9eb3bd4acc3836 as denominated

  • radv: make sure fp16 is enabled consistently on gfx8

Faith Ekstrand (2):

  • vulkan/queue: Move shared binary semaphores to temps

  • spirv: Add support for OpBitcast in OpSpecConstantOp

Georg Lehmann (1):

  • aco/lower_branches: update branch hints after changing jump targets

Hans-Kristian Arntzen (1):

  • radv/sqtt: Ensure that present fence gets signalled.

Hyunjun Ko (1):

  • vulkan/video: fix misuse of CLAMP in h265 slice parsing.

Ian Romanick (3):

  • brw: Increase the size of some structure fields in combine_constants

  • elk: Increase the size of some structure fields in combine_constants

  • brw/nir: nir_intrinsic_load_reloc_const_intel may not be scalar [v3]

Jianxun Zhang (1):

  • iris: Disable compression on sharing without modifier

Jordan Justen (1):

  • anv: Use image view base-layer in can_fast_clear_color_att()

José Roberto de Souza (1):

  • intel/brw: Use ASR over SHR for SHADER_OPCODE_ISUB_SAT

Karol Herbst (1):

  • Revert “ci: Update CrosVM and Virglrenderer”

Kenneth Graunke (2):

  • nir: Add load_simd_width_intel to divergence analysis

  • intel/nir: Make ffma peephole optimization preserve fp_fast_math flags

Lucas Fryzek (1):

  • anv: Update viewport/scissor state when count changes

Mario Kleiner (2):

  • asahi: Fix lseek failure error handling in agx_bo_import().

  • asahi: Set PIPE_BIND_SCANOUT in agx_resource_from_handle().

Mel Henning (2):

  • nak: Don’t copy-prop adds that flush to zero

  • nak: Fix divergence test for redux availability

Mike Blumenkrantz (7):

  • zink: wait on queues during screen destroy

  • zink: account for kopper dt not having a swapchain when pruning batch usage

  • zink: check for zink_batch_state::ctx before using during descriptor state reset

  • zink: flag gfx pipeline_changed if switching from a shader object draw

  • tu: don’t deref end info in tu_CmdEndRendering2EXT

  • mesa: don’t assert when finding a renderbuffer miplevel fails

  • zink: fix u_blitting when clears are pending

Nanley Chery (8):

  • anv: Disable CCS if image bound to wrong heap on Xe2+

  • anv: Disable fast-clears on linear surfaces

  • iris: Disable fast-clears on linear surfaces

  • iris: Add PIPE_BIND_SCANOUT when exporting textures

  • iris: Fix image reallocation for sharing

  • intel/isl: Only set CMF on renderable views on Xe2+

  • intel: Enable CCS_E on linear surfaces on Xe2+

  • iris: Drop iris_resource_image_is_pat_compressible

Olivia Lee (2):

  • pan/va: fix bi_is_imm_desc_handle early return

  • panvk: fix FS driver set layout when LD_VAR_BUF is disabled

Qiang Yu (1):

  • radeonsi: fix use aco/llvm debug options

Rhys Perry (2):

  • aco: workaround load tearing for load_shared2_amd

  • aco: fix SGPR 8-bit nir_op_vec with mixed constant and non-constant

Tapani Pälli (1):

  • blorp: add missing pipecontrol after 3DSTATE_WM_HZ_OP for Xe2+

Timothy Arceri (1):

  • nir: fix uniform cloning helper again

Timur Kristóf (2):

  • spirv: Always mark FS layer and viewport index inpus as flat

  • ac/nir/ngg: Fix scalarized mesh primitive indices