gh-81057: Add a CI Check for New Unsupported C Global Variables (gh-102506)

This will keep us from adding new unsupported (i.e. non-const) C global variables, which would break interpreter isolation.

FYI, historically it is very uncommon for new global variables to get added. Furthermore, it is rare for new code to break the c-analyzer. So the check should almost always pass unnoticed.

Note that I've removed test_check_c_globals. A test wasn't a great fit conceptually and was super slow on debug builds. A CI check is a better fit.

This also resolves gh-100237.

https://github.com/python/cpython/issues/81057

2 files changed, 19 insertions, 10 deletions

diff --git a/Tools/c-analyzer/c_parser/preprocessor/common.py b/Tools/c-analyzer/c_parser/preprocessor/common.py
index 4291a06..dbe1ede 100644
--- a/Tools/c-analyzer/c_parser/preprocessor/common.py
+++ b/Tools/c-analyzer/c_parser/preprocessor/common.py
@@ -115,15 +115,15 @@ def converted_error(tool, argv, filename):
 def convert_error(tool, argv, filename, stderr, rc):
     error = (stderr.splitlines()[0], rc)
     if (_expected := is_os_mismatch(filename, stderr)):
-        logger.debug(stderr.strip())
+        logger.info(stderr.strip())
         raise OSMismatchError(filename, _expected, argv, error, tool)
     elif (_missing := is_missing_dep(stderr)):
-        logger.debug(stderr.strip())
+        logger.info(stderr.strip())
         raise MissingDependenciesError(filename, (_missing,), argv, error, tool)
     elif '#error' in stderr:
         # XXX Ignore incompatible files.
         error = (stderr.splitlines()[1], rc)
-        logger.debug(stderr.strip())
+        logger.info(stderr.strip())
         raise ErrorDirectiveError(filename, argv, error, tool)
     else:
         # Try one more time, with stderr written to the terminal.
diff --git a/Tools/c-analyzer/c_parser/preprocessor/gcc.py b/Tools/c-analyzer/c_parser/preprocessor/gcc.py
index 7ef1a8a..24c1b0e 100644
--- a/Tools/c-analyzer/c_parser/preprocessor/gcc.py
+++ b/Tools/c-analyzer/c_parser/preprocessor/gcc.py
@@ -6,6 +6,11 @@ from . import common as _common
 
 TOOL = 'gcc'
 
+META_FILES = {
+    '<built-in>',
+    '<command-line>',
+}
+
 # https://gcc.gnu.org/onlinedocs/cpp/Preprocessor-Output.html
 # flags:
 #  1  start of a new file
@@ -75,11 +80,15 @@ def _iter_lines(text, reqfile, samefiles, cwd, raw=False):
 
     # The first line is special.
     # The next two lines are consistent.
-    for expected in [
-        f'# 1 "{reqfile}"',
-        '# 1 "<built-in>"',
-        '# 1 "<command-line>"',
-    ]:
+    firstlines = [
+        f'# 0 "{reqfile}"',
+        '# 0 "<built-in>"',
+        '# 0 "<command-line>"',
+    ]
+    if text.startswith('# 1 '):
+        # Some preprocessors emit a lineno of 1 for line-less entries.
+        firstlines = [l.replace('# 0 ', '# 1 ') for l in firstlines]
+    for expected in firstlines:
         line = next(lines)
         if line != expected:
             raise NotImplementedError((line, expected))
@@ -121,7 +130,7 @@ def _iter_top_include_lines(lines, topfile, cwd,
     # _parse_marker_line() that the preprocessor reported lno as 1.
     lno = 1
     for line in lines:
-        if line == '# 1 "<command-line>" 2':
+        if line == '# 0 "<command-line>" 2' or line == '# 1 "<command-line>" 2':
             # We're done with this top-level include.
             return
 
@@ -174,8 +183,8 @@ def _parse_marker_line(line, reqfile=None):
         return None, None, None
     lno, origfile, flags = m.groups()
     lno = int(lno)
+    assert origfile not in META_FILES, (line,)
     assert lno > 0, (line, lno)
-    assert origfile not in ('<built-in>', '<command-line>'), (line,)
     flags = set(int(f) for f in flags.split()) if flags else ()
 
     if 1 in flags: