libvpx yellow billed duck v1.17.0-rc2

[Jerome Jiang]

Hello,

Yellow Billed Duck v1.17.0-rc2 has been tagged[1]. Please help testing and report issues[2].
This release is ABI incompatible with the previous release. This release candidate includes more bug fixes.

Thanks.

[1]https://chromium.googlesource.com/webm/libvpx/+/refs/tags/v1.17.0-rc2
[2]https://g-issues.webmproject.org/issues/new?component=1615215&template=2023762

Changelog:

2026-08-07 v1.17.0 "Yellowbilled Duck"
  This release includes a new encoder control to validate high bitdepth input,
  additional Neon and AVX2/AVX512 optimizations, support for the C11 standard,
  and numerous bug fixes.

  - Upgrading:
    This release is ABI compatible with the previous release.
    ISO C11 standard is now enforced for C files.
    SVC is now strictly disallowed for 2-pass encoding; enabling it will return
    VPX_CODEC_INVALID_PARAM.
    VP9 High Bitdepth (HBD) input validation is now enabled by default. If the
    input contains values outside the valid range, the encoder will return
    VPX_CODEC_INVALID_PARAM. Use the new VP9E_SET_VALIDATE_HBD_INPUT control
    (or vpxenc --validate-hbd-input=0) to disable this validation.
    VP9 now sets the returned image width (w) and height (h) to the actual
    width and height, rather than the stride and aligned height. This is a
    behavior change that may affect applications relying on w being equal to the
    stride or h being the aligned height.
    Strict validation has been added to some APIs:
    - Encoders (VP8/VP9) now reject input images with unequal U and V strides.
    - vpx_img_alloc() and vpx_img_wrap() now strictly validate the image format.
    - External rate control (RTC) API now validates configuration inputs.
    - VP9 SVC now validates spatial layer ID input.

  - Enhancement:
    Added Arm Neon DotProd subpel variance paths.
    Optimized Neon HBD sadx4d and sad_skipx4d functions.
    Added AVX2/AVX512 implementations for various vpx_highbd predictor
    functions.
    Added support for MSVC v145 (Visual Studio 2026).
    Added configure option --disable-x86-asm.

  - Bug fixes:
    Fix to heap buffer overflow / out-of-bounds write in vp9_postproc,
    vp8-multi-res-encoding, vpx_setup_noise, set_mb_ssim_rdmult_scaling,
    vp9_change_config (resolution change), vp9_dec_alloc_row_mt_mem (decode
    coeffs), and vp8_sixtap_predict (LoongArch).
    Fix to integer overflow in vp9-svc (layer resolution, duration),
    set_mb_ssim_rdmult_scaling, vp8_encode_frame, calc_pframe_target_size,
    y4minput buffer size calculations, and get_token_alloc.
    Fix to VP8 decoder crash on flush with no fragments.
    Fix to VP8 decoder potential hang/slowdown with corrupt partition 0 size.
    Fix to VP9 decoder crash with invalid reference frame index.
    Fix to VP9 highbitdepth encoder format mismatch.
    Fix to VP9 float-to-int overflow in validate_config (EOS count) and with
    zero bitrate.
    Fix to VP8 decoder multithreading corruption flag reset at start of
    frame.
    Fix to VP8 decoder accepting inter frames on resolution change before
    keyframe.
    Fix to VP8 encoder reference frame dimension validation.
    Fix to VP8 encoder temporal layer ID validation.
    Fix to VP8 encoder missing setjmp in VP8E_SET_SCALEMODE.
    Fix to VP9 SVC encoder crash with dynamic resize and cyclic refresh
    (aq_mode=3).
    Fix to VP9 1-pass rate control last_boosted_qindex clamping and
    VP9E_SET_QUANTIZER_ONE_PASS validation.
    Fix to NV12 x_chroma_shift calculation in vpx_image and image2yuvconfig.
    Fix to AVX-512 CPU feature detection on older x86 CPUs.
    Fix to x86_abi_support.asm placing read-only data tables in .rdata for
    Windows.
    Fix to file leaks on error paths in vpxdec and video reader/writer
    utilities.
    Fix to crash/undefined behavior in vpx_img_flip with formats without alpha.