| Project | Fix | PR |
|---|---|---|
| Microsoft / DirectXMath | Fix strict aliasing in SSE packed vector stores |  |
| Apache Software Foundation / sedona | Interpolate ST_AddMeasure using cumulative lengths | |
| Apache Software Foundation / arrow-nanoarrow | Preserve half-float subnormal values | |
| Alibaba / MNN | Honor ONNX Resize's default nearest mode | |
| RISC-V / riscv-opcodes | Fix Rust masks for vector segment loads and stores | |
| LLVM / circt | Only narrow signed subtraction in SubOfPad canonicalizers | |
| Khronos Group / KTX-Software | Fix RGB9E5 clamping in ktx create | |
| Khronos Group / KTX-Software-CTS | Update RGB9E5 golden files for the shared-exponent clamp fix | |
| WebAssembly / wabt | Fix memory indices on SIMD splat and zero loads | |
| Keras | Fix batched vector right-hand sides in NumPy and JAX solve | |
| Open Neural Network Exchange / optimizer | Preserve indices shape when eliminating Shape/Gather | |
| Arm Software / astc-encoder | Fix HDR decompression example in README | |
| Boost.org / url | Preserve plus signs in params_ref::set | |
| CHIPS Alliance / sv-tests | Fix unary operator simulation test metadata | |
| dmlc / xgboost | Fix exact interaction constraints with fewer rows than features | |
| enjoy-digital / litesata | Preserve bus data through Mem2Sector conversion | |
| enjoy-digital / litesata | Handle Device Fault on write completion | |
| Academy Software Foundation / openexr | Fix B44/B44A raw fallback when compressed size equals raw size | |
| RISC-V Software / riscv-tests | Initialize the full CBO.ZERO test block | |
| RISC-V Software / riscv-tests | Add one-way Zfh format-conversion tests | |
| radare org / radare2 | Use the overflow-checked array allocators in RUtil and RBin ##util | |
| FEniCS Project / basix | Fix interpolation for reordered Lagrange elements | |
| Abseil / abseil-cpp | Honor zero padding in slow scientific conversions | |
| Flatiron Institute / finufft | Handle empty source sets | |
| Verilator | Add case item extension coverage (#8720) | |
| serde-rs / json | Preserve arbitrary precision in Number deserialization | |
| Scikit-HEP Project / awkward | Preserve argmin and argmax positions after missing lists | |
| MessagePack / msgpack-c | Guard C++ buffer and zone size arithmetic against wraparound | |
| BORG Lab / gtsam | Fix robust weight queries to use whitened residuals | |
| Pillow | Allow floating point constants in ImageMath operations | |
| OpenMDAO | Fix BalanceComp constructor RHS keyword forwarding | |
| Unit8 SA / darts | Fix backtest ignoring val_length when generating forecasts | |
| gem5 | NaN-box finite fminm.s and fmaxm.s results | |
| CESNET / libyang | Plugins types REFACTOR build strings with ly_append_str | |
| ml.js / matrix | Use absolute values for max norm | |
| Sonos, Inc. / tract | Preserve truncation in signed integer division | |
| CasADi | Fix copysign derivatives for negative magnitude inputs | |
| ghdl | Preserve sign of integers with leading zeros | |
| FontTools | Inherit CFF2 vsindex from Private DICT | |
| Nyx Space / hifitime | Fix calendar reference in GNSS MJD initializers | |
| heitzmann / gdstk | Fix RobustPath half-width in OASIS output | |
| Open Source RF Engineering / scikit-rf | Fix splitter s_def for complex port impedances | |
| foliojs / fontkit | Fix extra quadratic segments at off-curve contour starts | |
| PBS / pycaption | Preserve DFXP captions from divisions with the same language | |
| JuliaApproximation / FastTransforms.jl | Fix highest coefficient normalization in Chebyshev U transforms | |
| Chemistry Development Kit / cdk | Fix TPSA contributions for sulfur and phosphorus in three-membered rings | |
| Julia DSP / DSP.jl | Fix filtfilt state for unnormalized IIR coefficients |
|
| Nickel Lang | Preserve large integers in nickel convert | |
| ceres-solver | Fix covariance truncation for wide Jacobians | |
| zlib-ng / minizip-ng | Read each extra field from a stream bounded to its length | |
| Tskit developers | Fix integer overflow in Tajimas_D | |
| veryl-lang | Use blocking assignments for local concatenations | |
| unicode-rs / unicode-segmentation | Reset pending requests when moving GraphemeCursor | |
| hyperrealm / libconfig | Use checked size arithmetic for internal container growth | |
| FormatJS | Keep fraction digits when selecting | |
| hdl-modules | Fix read data stability in axi_lite_register_file | |
| SanDisk Open Source / pyvcd | Format realtime variables as real values | |
| SanDisk Open Source / pyvcd | Use total width for compound vector aliases | |
| nodatime | Badíʿ PlusMonths for additions ending in month 19 | |
| flintlib | Fix exponent aliasing in arb_pow | |
| stillwater-sc / universal | Preserve exact integers when converting to int | |
| ispc | Fix unsigned 64-bit saturating multiplication overflow | |
| OpenHW Foundation / core-v-mcu | Preserve full uDMA I2C repeat count | |
| OpenHW Foundation / cv-hpdcache | Expire open write-buffer entries when the threshold becomes zero | |
| libexif Project | Share one range check and size multiply helper between the loaders | |
| opencollab / arpack-ng | Preserve default tolerance across reverse communication calls | |
| opencollab / arpack-ng | Restore NP before the 'BE' swap (fixes #499) | |
| ParmEd | Preserve explicit exclusions during GROMACS molecule assembly | |
| TagLib | Return excess values from setProperties | |
| Moqups / culori | Fix negative saturation in modern HSL parsing | |
| fribidi | Reset isolates at paragraph boundaries in direction scan | |
| Optimatika / ojAlgo | Fix binomial probabilities for large trial counts | |
| libical | icalcomponent.c - fix busy status of recurring events | |
