Fix the nightly link check: tell a broken runner from a dead link, repair every dead link it found (pytorch#21694)

Author: shoumikhin

.ci/docker/common/install_openssl.sh

@@ -5,8 +5,10 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-# NB: This script is needed for sccache and is adopted from PyTorch core repo at
-# https://github.com/pytorch/pytorch/blob/main/.ci/docker/common/install_openssl.sh
+# NB: This script is needed for sccache and is adopted from PyTorch core repo.
+# Core deleted its copy in pytorch/pytorch#179513, so this points at the last
+# revision that still had it instead of at a branch.
+# https://github.com/pytorch/pytorch/blob/9274b93eac9f026e9f51e282449b0300b8e05482/.ci/docker/common/install_openssl.sh
 set -ex
 
 OPENSSL=openssl-1.1.1k

.ci/scripts/test_llava.sh

@@ -114,7 +114,7 @@ export_llava() {
 # Download a new image
 download_image() {
     echo "Downloading image"
-    curl -o basketball.jpg https://upload.wikimedia.org/wikipedia/commons/7/73/Chicago_Bulls_and_New_Jersey_Nets%2C_March_28%2C_1991.jpg
+    curl -fL -o basketball.jpg https://upload.wikimedia.org/wikipedia/commons/3/3e/Chicago_Bulls_-_New_Jersey_Nets_match_on_March_28%2C_1991.jpg
 }
 
 run_and_verify() {

README.md

@@ -7,7 +7,7 @@
 <div align="center">
   <a href="https://pypi.org/project/executorch/"><img src="https://img.shields.io/pypi/v/executorch?style=for-the-badge&color=blue" alt="PyPI - Version"></a>
   <a href="https://github.com/pytorch/executorch/graphs/contributors"><img src="https://img.shields.io/github/contributors/pytorch/executorch?style=for-the-badge&color=blue" alt="GitHub - Contributors"></a>
-  <a href="https://github.com/pytorch/executorch/stargazers"><img src="https://img.shields.io/github/stars/pytorch/executorch?style=for-the-badge&color=blue" alt="GitHub - Stars"></a>
+  <a href="https://github.com/pytorch/executorch/stargazers"><!-- @lint-ignore GitHub serves 404 on /stargazers to non-browser clients --><img src="https://img.shields.io/github/stars/pytorch/executorch?style=for-the-badge&color=blue" alt="GitHub - Stars"></a>
   <a href="https://discord.gg/Dh43CKSAdc"><img src="https://img.shields.io/badge/Discord-Join%20Us-blue?logo=discord&logoColor=white&style=for-the-badge" alt="Discord - Chat with Us"></a>
   <a href="https://docs.pytorch.org/executorch/main/index.html"><img src="https://img.shields.io/badge/Documentation-blue?logo=googledocs&logoColor=white&style=for-the-badge" alt="Documentation"></a>
 </div>

backends/arm/scripts/toolchain_utils.sh

@@ -75,7 +75,9 @@ function musl_select_toolchain() {
     fi
 
     if [[ "${ARCH}" == "x86_64" ]] ; then
-        toolchain_url="https://musl.cc/aarch64-linux-musl-cross.tgz"
+        # musl.cc serves this fine to developers but has never answered the link
+        # check from a GitHub hosted runner.
+        toolchain_url="https://musl.cc/aarch64-linux-musl-cross.tgz" # @lint-ignore
         toolchain_dir="aarch64-linux-musl-cross"
         toolchain_md5_checksum="a6bb806af217a91cf575e15163e8b12b"
         toolchain_archive="${toolchain_dir}.tgz"

backends/qualcomm/scripts/download_qnn_sdk.py

@@ -75,7 +75,7 @@ def _read_qnn_config() -> Dict[str, str]:
 QNN_VERSION = _QNN_CONFIG.get("QNN_VERSION", "2.37.0.250724")
 QNN_ZIP_URL = _QNN_CONFIG.get(
     "QNN_ZIP_URL",
-    f"https://softwarecenter.qualcomm.com/api/download/software/sdks/"
+    f"https://softwarecenter.qualcomm.com/api/download/software/sdks/"  # @lint-ignore only half of a URL, the rest is on the next line
     f"Qualcomm_AI_Runtime_Community/All/{QNN_VERSION}/v{QNN_VERSION}.zip",
 )
 

docs/source/kernel-library-custom-aten-kernel.md

@@ -207,7 +207,7 @@ Unlike the YAML entry API, the C++ API only uses C++ macros `EXECUTORCH_LIBRARY`
 
 Please refer to [Custom Ops Best Practices](#custom-ops-api-best-practices) on which API to use.
 
-Similar to [`TORCH_LIBRARY`](https://pytorch.org/cppdocs/library.html#library_8h_1a0bd5fb09d25dfb58e750d712fc5afb84) in PyTorch, `EXECUTORCH_LIBRARY` takes the operator name and the C++ function name and register them into ExecuTorch runtime.
+Similar to [`TORCH_LIBRARY`](https://docs.pytorch.org/cppdocs/api/library/index.html) in PyTorch, `EXECUTORCH_LIBRARY` takes the operator name and the C++ function name and register them into ExecuTorch runtime.
 
 #### Prepare custom kernel implementation
 

docs/source/success-stories.md

@@ -66,7 +66,7 @@ PrivateMind delivers a fully private AI assistant using ExecuTorch's .pte format
 
 NimbleEdge successfully integrated ExecuTorch with its open-source DeliteAI platform to enable agentic workflows orchestrated in Python on mobile devices. The extensible ExecuTorch ecosystem allowed implementation of on-device optimization techniques leveraging contextual sparsity. ExecuTorch significantly accelerated the release of "NimbleEdge AI" for iOS, enabling models like Qwen 2.5 with tool calling support and achieving up to 30% higher transactions per second.
 
-[Visit →](https://nimbleedge.com) • [Blog →](https://www.nimbleedge.com/blog/meet-nimbleedge-ai-the-first-truly-private-on-device-assistant) • [iOS App →](https://apps.apple.com/in/app/nimbleedge-ai/id6746237456)
+[Visit →](https://nimbleedge.com) • [Blog →](https://www.nimbleedge.com/blog/meet-nimbleedge-ai-the-first-truly-private-on-device-assistant)
 :::
 
 ::::

examples/models/llama/runner/generation.py

@@ -74,7 +74,7 @@ def __init__(
         self.use_kv_cache = use_kv_cache
         self.tokenizer = get_tokenizer(tokenizer_path, tokenizer_config_path)
         self.device = device
-        # For some models like qwen, mismatch is acceptable: https://github.com/QwenLM/Qwen2.5/issues/466#issuecomment-2146759706
+        # For some models like qwen, mismatch is acceptable: https://github.com/QwenLM/Qwen3/issues/466#issuecomment-2146759706
         if vocab_size != self.tokenizer.n_words:
             print(
                 "Warning - given vocab_size in params is unequal to tokenizer vocab size."

scripts/lint_urls.sh

@@ -12,6 +12,7 @@ trap 'kill 0' SIGINT
 status=0
 green='\e[1;32m'; red='\e[1;31m'; cyan='\e[1;36m'; yellow='\e[1;33m'; reset='\e[0m'
 user_agent="Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36"
+egress_probe_url="https://api.github.com"
 max_jobs=10
 pids=()
 
@@ -26,28 +27,18 @@ while IFS=: read -r filepath url; do
       sleep 1
       code=$(curl -k -gsLm30 --retry 3 --retry-delay 3 --retry-connrefused -o /dev/null -w "%{http_code}" -r 0-0 -A "$user_agent" "$url") || code=000
     fi
-    if [ "$code" -lt 200 ] || [ "$code" -ge 400 ]; then
-      sleep 1
-      request_id=$(curl -sS -G -H 'Accept: application/json' \
-        --data-urlencode "host=$url" \
-        --data-urlencode "max_nodes=1" \
-        --data-urlencode "node=us3.node.check-host.net" \
-        https://check-host.net/check-http \
-        | jq -r .request_id) || request_id=""
-      if [ -n "$request_id" ]; then
-        sleep 5
-        for _ in {1..5}; do
-          new_code=$(curl -sS -H 'Accept: application/json' \
-            "https://check-host.net/check-result/$request_id" \
-            | jq -r -e '.[][0][3]') || new_code=000
-          [[ "$new_code" =~ ^[0-9]+$ ]] || new_code=000
-          if [ "$new_code" -ge 200 ] && [ "$new_code" -lt 400 ]; then
-            code=$new_code
-            break
-          fi
-          sleep 5
-        done
+    if [[ "$code" == "000" ]]; then
+      # No HTTP response at all. Usually a dead host, but a runner with no egress
+      # gets 000 for everything, and failing 2000 live links helps nobody. One
+      # probe of a host the runner already depends on tells the two apart.
+      probe=$(curl -k -gsLm10 -o /dev/null -w "%{http_code}" "$egress_probe_url") || probe=000
+      if [[ "$probe" == "000" ]]; then
+        printf "${yellow}WARN %s${reset} ${cyan}%s${reset} %s\n" "$code" "$url" "$filepath"
+        exit 0
       fi
+      # Egress works, so the host is either dead or slow. Give it one more try
+      # with a longer timeout before failing it.
+      code=$(curl -k -gsLm60 -o /dev/null -w "%{http_code}" -A "$user_agent" "$url") || code=000
     fi
     # Treat Cloudflare JS-challenge and rate-limit as success.
     if [[ "$code" == "403" || "$code" == "429" || "$code" == "503" ]]; then

setup.py

@@ -5,9 +5,10 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-# Part of this code is from pybind11 cmake_example:
-# https://github.com/pybind/cmake_example/blob/master/setup.py so attach the
-# license below.
+# Part of this code is from pybind11 cmake_example, so attach the license below.
+# That project has since dropped setup.py, so this points at the last revision
+# that still had it instead of at a branch.
+# https://github.com/pybind/cmake_example/blob/7a94877f581a14de4de1a096fb053a55fc2a66bf/setup.py
 
 # Copyright (c) 2016 The Pybind Development Team, All rights reserved.
 #