Michael/Loki
1
0 Fork 0
mirror of https://github.com/R0m1k3/Loki.git synced 2026-10-17 12:16:21 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
fe51bdeee903d2fffaf580ace737976d46641f30
Loki/rust/loki-rs/tests/parity_backends/build.json
T
Claude 5197ee2542 Lot 41 : compilation de llama.cpp depuis les sources tenue par la façade 
Avec LOKI_BACKENDS=rust, la façade compile elle-même llama.cpp sous Linux
et macOS (backends/build.rs) : install {force}, update {clean} et
install-custom {repo, name, ref}, port de backend_build.go, des jobs
lcRunInstall / lcRunUpdate / lcBuildAndSwitch et d'installCustomBackend.

- plan complet (Metal, CUDA avec racine du toolkit et architectures, HIP,
  Vulkan), build/ déplacé ou propre mis de côté en build.old ;
- étapes suivies : sorties des deux flux dans un même tube, découpage du
  sinkWriter (git) et du scanner (CMake), configure.log et build.log,
  fin des journaux remontée en cas d'échec ;
- outils manquants installés au mieux, compilateur vérifié ;
- service moteur arrêté le temps du build et relancé même en échec ;
- refus comme le Go : moteur de l'image, URL absente, optimisation en
  cours, job déjà lancé.

Côté Go, facadeBackendsRoutes gagne les trois routes hors Windows (la
chaîne MSVC reste au Go, qui rapporte ses étapes au job de la façade).

Parité : internal/loki/parity_build_test.go tire un journal de 15 étapes
(tests/parity_backends/build.json) que tests/backends.rs rejoue par la
façade ; tests unitaires du découpage, du plan, du cache et des erreurs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016tcGfJWefW5kUGoC2wL8cv
2026-10-16 04:54:32 +00:00

483 lines
18 KiB
JSON
Raw Blame History

{
"tools": {
"c++": "#!/bin/sh\n",
"cc": "#!/bin/sh\n",
"cmake": "#!/bin/sh\nif [ \"$1\" = \"-B\" ]; then\n mkdir -p build\n echo \"-- The C compiler identification is GNU 13.2.0\"\n echo \"-- Configuring done (0.5s)\"\n printf 'CMAKE_HOME_DIRECTORY:INTERNAL=%s\\n' \"$(pwd)\" \u003e build/CMakeCache.txt\n echo \"-- Build files have been written to: build\"\n exit 0\nfi\nif [ -f FAIL_BUILD ]; then\n echo \"[ 10%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o\"\n echo \"common/arg.cpp:12:5: error: 'boom' was not declared in this scope\"\n exit 2\nfi\necho \"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o\"\necho \"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o\"\necho \"[100%] Linking CXX executable ../bin/llama-server\"\nmkdir -p build/bin\nprintf '#!/bin/sh\\necho llama-server\\n' \u003e build/bin/llama-server\nchmod +x build/bin/llama-server\n",
"git": "#!/bin/sh\nd=$(dirname \"$0\")\ncase \"$1\" in\n clone)\n dir=\"$4\"\n if [ -f \"$(dirname \"$dir\")/CLONE_FAIL\" ]; then echo \"fatal: repository '$3' not found\" \u003e\u00262; exit 128; fi\n mkdir -p \"$dir/.git\"\n cat \"$d/remote-sha\" \u003e \"$dir/.git/HEAD_SHA\"\n echo \"Cloning into '$dir'...\" \u003e\u00262 ;;\n fetch)\n exit 0 ;;\n rev-parse)\n case \"$2\" in\n --short)\n if [ \"$3\" = HEAD ]; then head -c 7 .git/HEAD_SHA; else head -c 7 \"$d/remote-sha\"; fi\n echo ;;\n --abbrev-ref) echo master ;;\n HEAD) cat .git/HEAD_SHA ;;\n *) cat \"$d/remote-sha\" ;;\n esac ;;\n pull)\n cat \"$d/remote-sha\" \u003e .git/HEAD_SHA\n echo \"Fast-forward\" ;;\n checkout)\n echo \"$2\" \u003e .git/REF\n echo \"HEAD is now at $2\" \u003e\u00262 ;;\nesac\n",
"make": "#!/bin/sh\n"
},
"steps": [
{
"name": "mise à jour sans dépôt",
"remote": "aaaaaaa1111111111111111111111111111111",
"action": "update",
"job": {
"action": "update",
"compiled": 0,
"error": "aucun dépôt llama.cpp ({{HOME}}/backends/llama.cpp) — lance d'abord l'installation",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"✗ aucun dépôt llama.cpp ({{HOME}}/backends/llama.cpp) — lance d'abord l'installation"
],
"new": "",
"next": 2,
"old": "",
"phase": "échec",
"running": false
},
"bin": "",
"build_old": []
},
{
"name": "installation",
"action": "install",
"job": {
"action": "install",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ clone de llama.cpp…",
"Cloning into '{{HOME}}/backends/llama.cpp'...",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp/build/bin/llama-server",
"BIN mis à jour",
"✓ installation terminée"
],
"new": "",
"next": 14,
"old": "",
"phase": "installation terminée",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "mise à jour : déjà à jour",
"action": "update",
"job": {
"action": "update",
"compiled": 0,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ git fetch origin…",
"✓ déjà à jour (aaaaaaa) — rien à faire"
],
"new": "",
"next": 3,
"old": "aaaaaaa",
"phase": "déjà à jour (aaaaaaa) — rien à faire",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "mise à jour : nouveau commit",
"remote": "bbbbbbb2222222222222222222222222222222",
"action": "update",
"job": {
"action": "update",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ git fetch origin…",
"▶ git pull origin/master…",
"Fast-forward",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp/build/bin/llama-server",
"BIN mis à jour",
"✓ mis à jour : aaaaaaa → bbbbbbb"
],
"new": "bbbbbbb",
"next": 15,
"old": "aaaaaaa",
"phase": "mis à jour : aaaaaaa → bbbbbbb",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "recompilation propre",
"action": "update",
"clean": true,
"job": {
"action": "update",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ git fetch origin…",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp/build/bin/llama-server",
"BIN mis à jour",
"✓ recompilé (bbbbbbb)"
],
"new": "bbbbbbb",
"next": 13,
"old": "bbbbbbb",
"phase": "recompilé (bbbbbbb)",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": [
"llama.cpp"
]
},
{
"name": "installation sur un dépôt présent",
"remote": "ccccccc3333333333333333333333333333333",
"action": "install",
"job": {
"action": "install",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ dépôt déjà présent — bascule en mise à jour",
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ git fetch origin…",
"▶ git pull origin/master…",
"Fast-forward",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp/build/bin/llama-server",
"BIN mis à jour",
"✓ mis à jour : bbbbbbb → ccccccc"
],
"new": "ccccccc",
"next": 17,
"old": "bbbbbbb",
"phase": "mis à jour : bbbbbbb → ccccccc",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": [
"llama.cpp"
]
},
{
"name": "build en échec",
"remote": "ddddddd4444444444444444444444444444444",
"files": {
"backends/llama.cpp/FAIL_BUILD": "1"
},
"action": "update",
"job": {
"action": "update",
"compiled": 2,
"error": "compilation échouée: exit status 2",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ git fetch origin…",
"▶ git pull origin/master…",
"Fast-forward",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 10%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"common/arg.cpp:12:5: error: 'boom' was not declared in this scope",
"--- fin de configure.log ---",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"--- fin de build.log ---",
"[ 10%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"common/arg.cpp:12:5: error: 'boom' was not declared in this scope",
"✗ compilation échouée: exit status 2"
],
"new": "ddddddd",
"next": 19,
"old": "ccccccc",
"phase": "échec",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": [
"llama.cpp"
]
},
{
"name": "installation forcée",
"files": {
"backends/llama.cpp/FAIL_BUILD": "-"
},
"action": "install",
"force": true,
"job": {
"action": "install",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ suppression du dépôt existant (--force)…",
"▶ clone de llama.cpp…",
"Cloning into '{{HOME}}/backends/llama.cpp'...",
"plan de build : backend=cpu arch= jobs={{N}}",
"▶ configuration CMake…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp/build/bin/llama-server",
"BIN mis à jour",
"✓ installation terminée"
],
"new": "",
"next": 15,
"old": "",
"phase": "installation terminée",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "cmake absent",
"no_tools": [
"cmake"
],
"action": "update",
"job": {
"action": "update",
"compiled": 0,
"error": "outils toujours manquants après tentative d'install: cmake — installe-les à la main puis réessaie",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"✗ outils toujours manquants après tentative d'install: cmake — installe-les à la main puis réessaie"
],
"new": "",
"next": 2,
"old": "",
"phase": "échec",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "compilateur absent",
"no_tools": [
"cc",
"make"
],
"action": "update",
"job": {
"action": "update",
"compiled": 0,
"error": "compilateur C/C++ introuvable — installe gcc/g++/make (ou build-essential) manuellement",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"✗ compilateur C/C++ introuvable — installe gcc/g++/make (ou build-essential) manuellement"
],
"new": "",
"next": 2,
"old": "",
"phase": "échec",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "clone en échec",
"files": {
"backends/CLONE_FAIL": "1",
"backends/llama.cpp": "-"
},
"action": "install",
"job": {
"action": "install",
"compiled": 0,
"error": "git clone a échoué : exit status 128",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ clone de llama.cpp…",
"fatal: repository 'https://github.com/ggml-org/llama.cpp.git' not found",
"✗ git clone a échoué : exit status 128"
],
"new": "",
"next": 4,
"old": "",
"phase": "échec",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "custom : URL invalide",
"files": {
"backends/CLONE_FAIL": "-"
},
"action": "custom",
"url": "ftp://exemple/x",
"job": {
"action": "custom",
"compiled": 0,
"error": "URL de dépôt invalide (attendu https://…, git@… ou ssh://…) : ftp://exemple/x",
"exists": true,
"lines": [
"✗ URL de dépôt invalide (attendu https://…, git@… ou ssh://…) : ftp://exemple/x"
],
"new": "",
"next": 1,
"old": "",
"phase": "échec",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "custom : nouveau fork",
"action": "custom",
"url": "https://github.com/prism/llama.cpp.git",
"job": {
"action": "custom",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ clone de https://github.com/prism/llama.cpp.git…",
"Cloning into '{{HOME}}/backends/llama.cpp-prism'...",
"▶ compilation (backend=cpu)…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp-prism/build/bin/llama-server",
"→ pour l'utiliser : édite un modèle → section Moteur → « backend détecté » et choisis-le.",
"✓ backend custom installé"
],
"new": "",
"next": 13,
"old": "",
"phase": "backend custom installé",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "custom : nom assaini, sur un ref",
"action": "custom",
"url": "https://github.com/prism/llama.cpp.git",
"bname": "Mon Fork/../x",
"ref": "v2",
"job": {
"action": "custom",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ clone de https://github.com/prism/llama.cpp.git…",
"Cloning into '{{HOME}}/backends/mon-fork-..-x'...",
"HEAD is now at v2",
"▶ compilation (backend=cpu)…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/mon-fork-..-x/build/bin/llama-server",
"→ pour l'utiliser : édite un modèle → section Moteur → « backend détecté » et choisis-le.",
"✓ backend custom installé"
],
"new": "",
"next": 14,
"old": "",
"phase": "backend custom installé",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
},
{
"name": "custom : dossier existant sur un ref",
"action": "custom",
"url": "git@github.com:prism/llama.cpp.git",
"ref": "v3",
"job": {
"action": "custom",
"compiled": 2,
"error": "",
"exists": true,
"lines": [
"▶ vérification des outils (git, cmake, compilateur)…",
"▶ dépôt déjà présent — mise à jour…",
"HEAD is now at v3",
"▶ compilation (backend=cpu)…",
"-- The C compiler identification is GNU 13.2.0",
"-- Configuring done (0.5s)",
"-- Build files have been written to: build",
"[ 50%] Building CXX object common/CMakeFiles/common.dir/arg.cpp.o",
"[ 90%] Building CUDA object ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/fattn.cu.o",
"[100%] Linking CXX executable ../bin/llama-server",
"binaire compilé : {{HOME}}/backends/llama.cpp-prism/build/bin/llama-server",
"→ pour l'utiliser : édite un modèle → section Moteur → « backend détecté » et choisis-le.",
"✓ backend custom installé"
],
"new": "",
"next": 13,
"old": "",
"phase": "backend custom installé",
"running": false
},
"bin": "{{HOME}}/backends/llama.cpp/build/bin/llama-server",
"build_old": []
}
]
}
Reference in new issue View Git Blame Copy Permalink