diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index 237dac0..44c3bd5 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -6,7 +6,8 @@ permissions: contents: read jobs: publish: - runs-on: ubuntu-latest + # The API snapshot is clang's AST on ubuntu-24.04, where ci.yml audits it; ubuntu-latest moved to 26.04 (clang 22.1.2), where the audit reads a different AST. + runs-on: ubuntu-24.04 timeout-minutes: 30 permissions: contents: write diff --git a/README.md b/README.md index 0adaf4c..48d4c6a 100644 --- a/README.md +++ b/README.md @@ -19,14 +19,14 @@ ## Add The Package ```sh -mcpp new llama_app --template mcpplibs.mwrap.llamacpp@b10069.1 +mcpp new llama_app --template mcpplibs.mwrap.llamacpp@b10069.2 cd llama_app mcpp run ``` ```toml [dependencies] -mcpplibs.mwrap.llamacpp = "b10069.1" +mcpplibs.mwrap.llamacpp = "b10069.2" ``` ## Minimal CPU Program @@ -58,7 +58,7 @@ Shader generation, libstdc++ constraints and driver boundaries are in [usage](do ## Supported Boundary -Pin `b10069.1` exactly; it is not a semver range. `.macros` exposes mapped constants. +Pin `b10069.2` exactly; it is not a semver range. `.macros` exposes mapped constants. Model quality and unsupported GPU configurations are not implied by the CPU template. [Upstream policy](docs/upstream-update-policy.md) · [Validation](docs/validation/b10069.md) · [Apache-2.0](LICENSE) diff --git a/README.zh.md b/README.zh.md index 6d157fd..4c30568 100644 --- a/README.zh.md +++ b/README.zh.md @@ -19,14 +19,14 @@ ## 添加依赖 ```sh -mcpp new llama_app --template mcpplibs.mwrap.llamacpp@b10069.1 +mcpp new llama_app --template mcpplibs.mwrap.llamacpp@b10069.2 cd llama_app mcpp run ``` ```toml [dependencies] -mcpplibs.mwrap.llamacpp = "b10069.1" +mcpplibs.mwrap.llamacpp = "b10069.2" ``` ## 最小 CPU 程序 @@ -58,7 +58,7 @@ shader 生成、libstdc++ 限制及驱动边界见[使用说明](docs/usage.md) ## 支持边界 -精确固定 `b10069.1`,不使用 semver 范围;`.macros` 导出映射常量。 +精确固定 `b10069.2`,不使用 semver 范围;`.macros` 导出映射常量。 CPU 模板通过不代表模型质量或未支持 GPU 配置通过。 [上游策略](docs/upstream-update-policy.md) · [验证](docs/validation/b10069.md) · [Apache-2.0](LICENSE) diff --git a/docs/upstream-update-policy.md b/docs/upstream-update-policy.md index b7b6c4f..1aeb859 100644 --- a/docs/upstream-update-policy.md +++ b/docs/upstream-update-policy.md @@ -63,6 +63,9 @@ Every release records: - material API, model, and backend changes. Published tags, archives, and index entries are immutable. +`mwrap-b10069.1` was tagged but never released, because its release gate ran +on a changed runner image; `b10069.2` carries the same template change (the +dependency selector written unquoted). ## Cadence And Triggers diff --git a/docs/usage.md b/docs/usage.md index abbeade..dcfab01 100644 --- a/docs/usage.md +++ b/docs/usage.md @@ -14,7 +14,7 @@ revision of it: `b10069`. CPU is the default backend; Metal is an additive feature for macOS ARM64 and Vulkan an additive feature for Linux. Consumers pin exactly -- `b10069` is not semver, so no range expresses "b10069 or later" -- which means a consumer wanting a feature names the revision that has it. -`backend-vulkan` arrived in `b10069.1`. +`backend-vulkan` arrived in the legacy `ggml-org.llamacpp` `b10069.1`. ## Add The Package @@ -243,7 +243,7 @@ for exact evidence and remaining release gates. The b10069 package adds a runnable default template. After that version is indexed: ```sh -mcpp new my-llama --template mcpplibs.mwrap.llamacpp@b10069.1 +mcpp new my-llama --template mcpplibs.mwrap.llamacpp@b10069.2 cd my-llama mcpp run ``` diff --git a/mcpp.toml b/mcpp.toml index 559dea4..20db2ce 100644 --- a/mcpp.toml +++ b/mcpp.toml @@ -2,7 +2,7 @@ [package] namespace = "mcpplibs.mwrap" name = "llamacpp" -version = "b10069.1" +version = "b10069.2" description = "C++23 module package for llama.cpp b10069" license = "Apache-2.0" authors = ["mcpp-community"]