Support mixing remote and local configs in EXTENDS

EXTENDS accepts both absolute URLs and relative file paths to config
files to inherit from. Stop assuming that all relative file paths are
relative to the current workspace. This assumption does not hold when
parsing a config inherited from a different repository. This situation
arises most simply when A inherits from B via an absolute URL, and B
inherits from C via a relative file path. Both inherited config files,
B and C, are in a different repository than A. Make a best effort to
infer the URL of the repository root, and use that to correctly resolve
relative file paths within that repository recursively.

Author: Kurt-von-Laven

.github/linters/.cspell.json

@@ -501,6 +501,7 @@
         "cmidrule",
         "codacy",
         "codebases",
+        "codeberg",
         "codeclimate",
         "codecov",
         "codenarcargs",
@@ -680,6 +681,8 @@
         "gijsreyn",
         "gitattributes",
         "gitblame",
+        "gitea",
+        "gitee",
         "gitlab",
         "gitleaks",
         "gitmodified",
@@ -733,6 +736,7 @@
         "htmlhint",
         "htmlhintrc",
         "htmlout",
+        "huggingface",
         "hyhs",
         "idiv",
         "ighe",
@@ -777,6 +781,7 @@
         "joereynolds",
         "jscoverage",
         "jscpd",
+        "jsdelivr",
         "jsonify",
         "jsonlint",
         "jsonlintrc",
@@ -1003,6 +1008,7 @@
         "packagename",
         "pagebreak",
         "pageref",
+        "pagure",
         "pandoc",
         "parallelization",
         "paren",
@@ -1016,6 +1022,7 @@
         "perlcriticrc",
         "pgfpicture",
         "phar",
+        "phcdn",
         "phive",
         "phpcs",
         "phplint",
@@ -1132,6 +1139,7 @@
         "returncode",
         "returnrules",
         "rexec",
+        "rhodecode",
         "risd",
         "rmfamily",
         "rockspec",

CHANGELOG.md

@@ -22,6 +22,9 @@ Note: Can be used with `oxsecurity/megalinter@beta` in your GitHub Action mega-l
   - Upgrade create-pull-request and create-or-update-comment GitHub Actions
   - Increase auto-update-linters GitHub Action timeout
   - Upgrade base Docker image to python:3.11.3-alpine3.17
+  - Fix a config inheritance bug that prevented extending a remote config that
+    extends a local config by @Kurt-von-Laven
+    ([#2371](https://github.com/oxsecurity/megalinter/issues/2371)).
 
 - Documentation
 

megalinter/config.py

@@ -4,13 +4,22 @@
 import os
 import shlex
 import tempfile
+from collections.abc import Mapping, Sequence
+from pathlib import Path, PurePath
+from typing import AnyStr, cast
+from urllib.parse import ParseResult, urlparse, urlunparse
 
 import requests
 import yaml
 
 CONFIG_DATA = None
 CONFIG_SOURCE = None
 
+JsonValue = (
+    None | bool | int | float | str | Sequence["JsonValue"] | Mapping[str, "JsonValue"]
+)
+JsonObject = dict[str, JsonValue]
+
 
 def init_config(workspace=None):
     global CONFIG_DATA, CONFIG_SOURCE
@@ -72,7 +81,7 @@ def init_config(workspace=None):
         )
     # manage EXTENDS in configuration
     if "EXTENDS" in runtime_config:
-        combined_config = {}
+        combined_config: JsonObject = {}
         CONFIG_SOURCE = combine_config(
             workspace, runtime_config, combined_config, CONFIG_SOURCE
         )
@@ -82,22 +91,32 @@ def init_config(workspace=None):
     set_config(runtime_config)
 
 
-def combine_config(workspace, config, combined_config, config_source):
-    extends = config["EXTENDS"]
+def combine_config(
+    workspace: str | None,
+    config: JsonObject,
+    combined_config: JsonObject,
+    config_source: str,
+    child_uri: ParseResult | None = None,
+) -> str:
+    workspace_path = Path(workspace) if workspace else None
+    parsed_uri: ParseResult | None = None
+    extends = cast(str | Sequence[str], config["EXTENDS"])
     if isinstance(extends, str):
         extends = extends.split(",")
     for extends_item in extends:
         if extends_item.startswith("http"):
-            r = requests.get(extends_item, allow_redirects=True)
-            assert (
-                r.status_code == 200
-            ), f"Unable to retrieve EXTENDS config file {extends_item}"
-            extends_config_data = yaml.safe_load(r.content)
+            parsed_uri = urlparse(extends_item)
+            extends_config_data = download_config(extends_item)
         else:
-            with open(
-                workspace + os.path.sep + extends_item, "r", encoding="utf-8"
-            ) as f:
-                extends_config_data = yaml.safe_load(f)
+            path = PurePath(extends_item)
+            if child_uri:
+                parsed_uri = resolve_uri(child_uri, path)
+                uri = urlunparse(parsed_uri)
+                extends_config_data = download_config(uri)
+            else:
+                resolved_path = workspace_path / path if workspace_path else Path(path)
+                with resolved_path.open("r", encoding="utf-8") as f:
+                    extends_config_data = yaml.safe_load(f)
         combined_config.update(extends_config_data)
         config_source += f"\n[config] - extends from: {extends_item}"
         if "EXTENDS" in extends_config_data:
@@ -106,11 +125,41 @@ def combine_config(workspace, config, combined_config, config_source):
                 extends_config_data,
                 combined_config,
                 config_source,
+                parsed_uri,
             )
     combined_config.update(config)
     return config_source
 
 
+def download_config(uri: AnyStr) -> JsonObject:
+    r = requests.get(uri, allow_redirects=True)
+    assert r.status_code == 200, f"Unable to retrieve EXTENDS config file {uri!r}"
+    return yaml.safe_load(r.content)
+
+
+def resolve_uri(child_uri: ParseResult, relative_config_path: PurePath) -> ParseResult:
+    match child_uri.netloc:
+        case "cdn.jsdelivr.net" | "git.launchpad.net":
+            repo_root_index = 3
+        case "code.rhodecode.com" | "git.savannah.gnu.org" | "raw.githubusercontent.com" | "repo.or.cz":
+            repo_root_index = 4
+        case "bitbucket.org" | "git.sr.ht" | "gitee.com" | "pagure.io":
+            repo_root_index = 5
+        case "codeberg.org" | "gitea.com" | "gitlab.com" | "huggingface.co" | "p.phcdn.net" | "sourceforge.net":
+            repo_root_index = 6
+        case _:
+            message = (
+                f"Unsupported Git repo hosting service: {child_uri.netloc}. "
+                "Request support be added to MegaLinter, or use absolute URLs "
+                "with EXTENDS in inherited configs rather than relative paths."
+            )
+            raise ValueError(message)
+    child_path = PurePath(child_uri.path)
+    repo_root_path = child_path.parts[:repo_root_index]
+    path = PurePath(*repo_root_path, str(relative_config_path))
+    return child_uri._replace(path=str(path))
+
+
 def get_config():
     global CONFIG_DATA
     if CONFIG_DATA is not None: