(Discussion/Proposals) Consistency regarding annotations of header-implemented functions (#316)

* Open discussion

* Move annotations of header-implemented functions back to `.h` files

* Adjust `README.md`

* Relocate annotation

* linter

* Comment markers in headers only, rename script, update github actions

* type hint compat

* Rename github action, better argparse for linter

* Type hints, working test for byname ignore

* Move annotation

* CI rename and enable warnfail, enforce mode always on

* Two step linting

* or one step

* continue on error

* two jobs instead

* Fixes

---------

Co-authored-by: disinvite <disinvite@users.noreply.github.com>

tools/isledecomp/isledecomp/parser/__init__.py

@@ -1 +1,2 @@
 from .parser import DecompParser
+from .linter import DecompLinter

tools/isledecomp/isledecomp/parser/error.py

@@ -1,6 +1,9 @@
 from enum import Enum
+from typing import Optional
+from dataclasses import dataclass
 
 
+# TODO: poorly chosen name, should be AlertType or AlertCode or something
 class ParserError(Enum):
     # WARN: Stub function exceeds some line number threshold
     UNLIKELY_STUB = 100
@@ -29,6 +32,16 @@ class ParserError(Enum):
     # and the start of the function. We can ignore it, but the line shouldn't be there
     UNEXPECTED_BLANK_LINE = 107
 
+    # WARN: We called the finish() method for the parser but had not reached the starting
+    # state of SEARCH
+    UNEXPECTED_END_OF_FILE = 108
+
+    # WARN: We found a marker to be referenced by name outside of a header file.
+    BYNAME_FUNCTION_IN_CPP = 109
+
+    # This code or higher is an error, not a warning
+    DECOMP_ERROR_START = 200
+
     # ERROR: We found a marker unexpectedly
     UNEXPECTED_MARKER = 200
 
@@ -39,3 +52,20 @@ class ParserError(Enum):
 
     # ERROR: The line following a synthetic marker was not a comment
     BAD_SYNTHETIC = 202
+
+    # ERROR: This function offset comes before the previous offset from the same module
+    # This hopefully gives some hint about which functions need to be rearranged.
+    FUNCTION_OUT_OF_ORDER = 203
+
+
+@dataclass
+class ParserAlert:
+    code: ParserError
+    line_number: int
+    line: Optional[str] = None
+
+    def is_warning(self) -> bool:
+        return self.code.value < ParserError.DECOMP_ERROR_START.value
+
+    def is_error(self) -> bool:
+        return self.code.value >= ParserError.DECOMP_ERROR_START.value

tools/isledecomp/isledecomp/parser/linter.py

@@ -0,0 +1,99 @@
+from typing import List, Optional
+from .parser import DecompParser
+from .error import ParserAlert, ParserError
+
+
+def get_checkorder_filter(module):
+    """Return a filter function on implemented functions in the given module"""
+    return lambda fun: fun.module == module and not fun.lookup_by_name
+
+
+class DecompLinter:
+    def __init__(self) -> None:
+        self.alerts: List[ParserAlert] = []
+        self._parser = DecompParser()
+        self._filename: str = ""
+        self._module: Optional[str] = None
+
+    def reset(self):
+        self.alerts = []
+        self._parser.reset()
+        self._filename = ""
+        self._module = None
+
+    def file_is_header(self):
+        return self._filename.lower().endswith(".h")
+
+    def _check_function_order(self):
+        """Rules:
+        1. Only markers that are implemented in the file are considered. This means we
+        only look at markers that are cross-referenced with cvdump output by their line
+        number. Markers with the lookup_by_name flag set are ignored because we cannot
+        directly influence their order.
+
+        2. Order should be considered for a single module only. If we have multiple
+        markers for a single function (i.e. for LEGO1 functions linked statically to
+        ISLE) then the virtual address space will be very different. If we don't check
+        for one module only, we would incorrectly report that the file is out of order.
+        """
+
+        if self._module is None:
+            return
+
+        checkorder_filter = get_checkorder_filter(self._module)
+        last_offset = None
+        for fun in filter(checkorder_filter, self._parser.functions):
+            if last_offset is not None:
+                if fun.offset < last_offset:
+                    self.alerts.append(
+                        ParserAlert(
+                            code=ParserError.FUNCTION_OUT_OF_ORDER,
+                            line_number=fun.line_number,
+                        )
+                    )
+
+            last_offset = fun.offset
+
+    def _check_offset_uniqueness(self):
+        # TODO
+        pass
+
+    def _check_byname_allowed(self):
+        if self.file_is_header():
+            return
+
+        for fun in self._parser.functions:
+            if fun.lookup_by_name:
+                self.alerts.append(
+                    ParserAlert(
+                        code=ParserError.BYNAME_FUNCTION_IN_CPP,
+                        line_number=fun.line_number,
+                    )
+                )
+
+    def check_lines(self, lines, filename, module=None):
+        """`lines` is a generic iterable to allow for testing with a list of strings.
+        We assume lines has the entire contents of the compilation unit."""
+
+        self.reset()
+        self._filename = filename
+        self._module = module
+
+        self._parser.read_lines(lines)
+
+        self._parser.finish()
+        self.alerts = self._parser.alerts[::]
+
+        if self._module is not None:
+            self._check_byname_allowed()
+            self._check_offset_uniqueness()
+
+            if not self.file_is_header():
+                self._check_function_order()
+
+        return len(self.alerts) == 0
+
+    def check_file(self, filename, module=None):
+        """Convenience method for decomplint cli tool"""
+        with open(filename, "r", encoding="utf-8") as f:
+            return self.check_lines(f, filename, module)

tools/isledecomp/isledecomp/parser/node.py

@@ -6,12 +6,6 @@ class ParserNode:
     line_number: int
 
 
-@dataclass
-class ParserAlert(ParserNode):
-    code: int
-    line: str
-
-
 @dataclass
 class ParserSymbol(ParserNode):
     module: str

tools/isledecomp/isledecomp/parser/parser.py

@@ -12,12 +12,11 @@ from .util import (
     remove_trailing_comment,
 )
 from .node import (
-    ParserAlert,
     ParserFunction,
     ParserVariable,
     ParserVtable,
 )
-from .error import ParserError
+from .error import ParserAlert, ParserError
 
 
 class ReaderState(Enum):
@@ -29,6 +28,7 @@ class ReaderState(Enum):
     IN_GLOBAL = 5
     IN_FUNC_GLOBAL = 6
     IN_VTABLE = 7
+    DONE = 100
 
 
 def marker_is_stub(marker: DecompMarker) -> bool:
@@ -56,7 +56,7 @@ def marker_is_vtable(marker: DecompMarker) -> bool:
 
 
 class MarkerDict:
-    def __init__(self):
+    def __init__(self) -> None:
         self.markers: dict = {}
 
     def insert(self, marker: DecompMarker) -> bool:
@@ -80,7 +80,7 @@ class DecompParser:
     # pylint: disable=too-many-instance-attributes
     # Could combine output lists into a single list to get under the limit,
     # but not right now
-    def __init__(self):
+    def __init__(self) -> None:
         # The lists to be populated as we parse
         self.functions: List[ParserFunction] = []
         self.vtables: List[ParserVtable] = []
@@ -306,6 +306,9 @@ class DecompParser:
             self._syntax_warning(ParserError.BOGUS_MARKER)
 
     def read_line(self, line: str):
+        if self.state == ReaderState.DONE:
+            return
+
         self.last_line = line  # TODO: Useful or hack for error reporting?
         self.line_number += 1
 
@@ -392,3 +395,9 @@ class DecompParser:
     def read_lines(self, lines: Iterable):
         for line in lines:
             self.read_line(line)
+
+    def finish(self):
+        if self.state != ReaderState.SEARCH:
+            self._syntax_warning(ParserError.UNEXPECTED_END_OF_FILE)
+
+        self.state = ReaderState.DONE