Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package llamacpp for openSUSE:Factory checked in at 2026-11-08 19:21:52 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/llamacpp (Old) and /work/SRC/openSUSE:Factory/.llamacpp.new.3759290 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "llamacpp" Sun Nov 8 19:21:52 2026 rev:40 rq:1396093 version:0.9.0 Changes: -------- --- /work/SRC/openSUSE:Factory/llamacpp/llamacpp.changes 2026-10-30 18:18:21.286474454 +0100 +++ /work/SRC/openSUSE:Factory/.llamacpp.new.3759290/llamacpp.changes 2026-11-08 19:22:19.017703011 +0100 @@ -1,0 +2,19 @@ +Sun Nov 08 00:00:00 UTC 2026 - Eyad Issa <[email protected]> + +- Update to version 0.9.0: + * Update ggml to 0.29.0 and add FP8 E4M3 quantization. + * Add Clover, Limite, Metis-1.6 and Mistral Large 4 models. + * Improve multi-sequence inference, audio support and Gemma 4 + prefill performance. + * Update the server web UI to build b12110. + +------------------------------------------------------------------- +Mon Nov 02 20:19:38 UTC 2026 - Eyad Issa <[email protected]> + +- Obsolete the discontinued libllava package with libmtmd0. +- Keep dynamically loaded ggml backends matched to the core + libraries. +- Recommend HIP and OpenVINO for AMD and Intel graphics devices, + respectively, instead of recommending them on every system. + +------------------------------------------------------------------- @@ -45,2 +64,4 @@ - EmbeddingGemma-2, LiquidAI d1/d1-omni, and vLLM decision models. - * Refactor server chat layer around common_chat_session and change + EmbeddingGemma-2, LiquidAI d1/d1-omni, and vLLM decision + models. + * Refactor server chat layer around common_chat_session and + change @@ -53,0 +75,6 @@ +- CVE-2026-116829: Fix integer overflow in GGUF tensor element + counts that could lead to out-of-bounds access (bsc#1288068). + * Fixed upstream in version 0.6.0. +- CVE-2026-116840: Fix integer overflow when padding GGUF tensor + sizes, which could lead to out-of-bounds reads. + * Fixed upstream in version 0.6.0. @@ -491,0 +519,3 @@ +- CVE-2026-33298: Fix integer overflow in GGUF tensor parsing that + could cause a heap buffer overflow (bsc#1260368). + * Fixed upstream in b7824. Old: ---- llama-b11960-ui.tar.gz llamacpp-0.8.0.tar.gz New: ---- llama-b12110-ui.tar.gz llamacpp-0.9.0.tar.gz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ llamacpp.spec ++++++ --- /var/tmp/diff_new_pack.gKpzXa/_old 2026-11-08 19:22:20.165750503 +0100 +++ /var/tmp/diff_new_pack.gKpzXa/_new 2026-11-08 19:22:20.167750586 +0100 @@ -18,7 +18,7 @@ %global backend_dir %{_libdir}/ggml -%global upstream_build 11960 +%global upstream_build 12110 %global license_dir %{_datadir}/licenses %global llama_sover %{version} @@ -27,8 +27,12 @@ %global mtmd_sover %{llama_sover} %global mtmd_sover_suffix 0 -%global ggml_sover 0.28.0 +%global ggml_sover 0.29.0 %global ggml_sover_suffix 0 +# PCI display controllers (class 03) from AMD and Intel, respectively. +# These broad matches do not guarantee backend support for every device. +%global amd_gpu_modalias modalias(pci:v00001002d*sv*sd*bc03sc*i*) +%global intel_gpu_modalias modalias(pci:v00008086d*sv*sd*bc03sc*i*) %if 0%{?suse_version} == 1500 %bcond_with opencl @@ -53,7 +57,7 @@ %endif Name: llamacpp -Version: 0.8.0 +Version: 0.9.0 Release: 0 Summary: Inference of Meta's LLaMA model (and others) in pure C/C++ License: Apache-2.0 AND MIT AND BSD-2-Clause AND BSD-3-Clause AND ISC AND MPL-2.0 @@ -141,14 +145,11 @@ %package -n libggml%{ggml_sover_suffix} Summary: A tensor library for C++ -Requires: libggml-cpu +Requires: libggml-cpu = %{version}-%{release} %if %{with opencl} -Recommends: libggml-opencl +Recommends: libggml-opencl = %{version}-%{release} %endif -%if %{with hip} -Recommends: libggml-hip -%endif -Recommends: libggml-vulkan +Recommends: libggml-vulkan = %{version}-%{release} %description -n libggml%{ggml_sover_suffix} A tensor library for C++. It was created originally to support llama.cpp @@ -165,6 +166,8 @@ %package -n libggml-cpu Summary: A tensor library for C++ (CPU backend) +Requires: libggml%{ggml_sover_suffix} = %{version}-%{release} +Requires: libggml-base%{ggml_sover_suffix} = %{version}-%{release} %description -n libggml-cpu A tensor library for C++. It was created originally to support llama.cpp @@ -174,6 +177,8 @@ %package -n libggml-vulkan Summary: A tensor library for C++ (Vulkan backend) +Requires: libggml%{ggml_sover_suffix} = %{version}-%{release} +Requires: libggml-base%{ggml_sover_suffix} = %{version}-%{release} %description -n libggml-vulkan A tensor library for C++. It was created originally to support llama.cpp @@ -184,6 +189,8 @@ %if %{with opencl} %package -n libggml-opencl Summary: A tensor library for C++ (OpenCL backend) +Requires: libggml%{ggml_sover_suffix} = %{version}-%{release} +Requires: libggml-base%{ggml_sover_suffix} = %{version}-%{release} %description -n libggml-opencl A tensor library for C++. It was created originally to support llama.cpp @@ -195,6 +202,9 @@ %if %{with hip} %package -n libggml-hip Summary: A tensor library for C++ (HIP/ROCm backend) +Requires: libggml%{ggml_sover_suffix} = %{version}-%{release} +Requires: libggml-base%{ggml_sover_suffix} = %{version}-%{release} +Supplements: (libggml%{ggml_sover_suffix} and %{amd_gpu_modalias}) %description -n libggml-hip A tensor library for C++. It was created originally to support llama.cpp @@ -206,6 +216,9 @@ %if %{with openvino} %package -n libggml-openvino Summary: A tensor library for C++ (OpenVINO backend) +Requires: libggml%{ggml_sover_suffix} = %{version}-%{release} +Requires: libggml-base%{ggml_sover_suffix} = %{version}-%{release} +Supplements: (libggml%{ggml_sover_suffix} and %{intel_gpu_modalias}) %description -n libggml-openvino A tensor library for C++. It was created originally to support llama.cpp @@ -228,6 +241,7 @@ %package -n libmtmd%{mtmd_sover_suffix} Summary: Library to run multimodals inference models +Obsoletes: libllava < 5402 %description -n libmtmd%{mtmd_sover_suffix} As outlined in the history, libmtmd is the modern library designed to ++++++ llama-b11960-ui.tar.gz -> llama-b12110-ui.tar.gz ++++++ ++++ 8137 lines of diff (skipped) ++++++ llamacpp-0.8.0.tar.gz -> llamacpp-0.9.0.tar.gz ++++++ /work/SRC/openSUSE:Factory/llamacpp/llamacpp-0.8.0.tar.gz /work/SRC/openSUSE:Factory/.llamacpp.new.3759290/llamacpp-0.9.0.tar.gz differ: char 13, line 1
