1 files changed, 84 insertions, 26 deletions

diff --git a/bitbake/lib/bb/codeparser.py b/bitbake/lib/bb/codeparser.py
index 25a7ac69d3..2e8b7ced3c 100644
--- a/bitbake/lib/bb/codeparser.py
+++ b/bitbake/lib/bb/codeparser.py
@@ -1,4 +1,6 @@
 #
+# Copyright BitBake Contributors
+#
 # SPDX-License-Identifier: GPL-2.0-only
 #
 
@@ -25,6 +27,7 @@ import ast
 import sys
 import codegen
 import logging
+import inspect
 import bb.pysh as pysh
 import bb.utils, bb.data
 import hashlib
@@ -56,10 +59,40 @@ def check_indent(codestr):
 
     return codestr
 
-# A custom getstate/setstate using tuples is actually worth 15% cachesize by
-# avoiding duplication of the attribute names!
+modulecode_deps = {}
 
+def add_module_functions(fn, functions, namespace):
+    import os
+    fstat = os.stat(fn)
+    fixedhash = fn + ":" + str(fstat.st_size) +  ":" + str(fstat.st_mtime)
+    for f in functions:
+        name = "%s.%s" % (namespace, f)
+        parser = PythonParser(name, logger)
+        try:
+            parser.parse_python(None, filename=fn, lineno=1, fixedhash=fixedhash+f)
+            #bb.warn("Cached %s" % f)
+        except KeyError:
+            lines, lineno = inspect.getsourcelines(functions[f])
+            src = "".join(lines)
+            parser.parse_python(src, filename=fn, lineno=lineno, fixedhash=fixedhash+f)
+            #bb.warn("Not cached %s" % f)
+        execs = parser.execs.copy()
+        # Expand internal module exec references
+        for e in parser.execs:
+            if e in functions:
+                execs.remove(e)
+                execs.add(namespace + "." + e)
+        modulecode_deps[name] = [parser.references.copy(), execs, parser.var_execs.copy(), parser.contains.copy()]
+        #bb.warn("%s: %s\nRefs:%s Execs: %s %s %s" % (name, fn, parser.references, parser.execs, parser.var_execs, parser.contains))
+
+def update_module_dependencies(d):
+    for mod in modulecode_deps:
+        excludes = set((d.getVarFlag(mod, "vardepsexclude") or "").split())
+        if excludes:
+            modulecode_deps[mod] = [modulecode_deps[mod][0] - excludes, modulecode_deps[mod][1] - excludes, modulecode_deps[mod][2] - excludes, modulecode_deps[mod][3]]
 
+# A custom getstate/setstate using tuples is actually worth 15% cachesize by
+# avoiding duplication of the attribute names!
 class SetCache(object):
     def __init__(self):
         self.setcache = {}
@@ -152,12 +185,12 @@ class CodeParserCache(MultiProcessCache):
         self.shellcachelines[h] = cacheline
         return cacheline
 
-    def init_cache(self, d):
+    def init_cache(self, cachedir):
         # Check if we already have the caches
         if self.pythoncache:
             return
 
-        MultiProcessCache.init_cache(self, d)
+        MultiProcessCache.init_cache(self, cachedir)
 
         # cachedata gets re-assigned in the parent
         self.pythoncache = self.cachedata[0]
@@ -169,8 +202,8 @@ class CodeParserCache(MultiProcessCache):
 
 codeparsercache = CodeParserCache()
 
-def parser_cache_init(d):
-    codeparsercache.init_cache(d)
+def parser_cache_init(cachedir):
+    codeparsercache.init_cache(cachedir)
 
 def parser_cache_save():
     codeparsercache.save_extras()
@@ -195,6 +228,10 @@ class BufferedLogger(Logger):
                 self.target.handle(record)
         self.buffer = []
 
+class DummyLogger():
+    def flush(self):
+        return
+
 class PythonParser():
     getvars = (".getVar", ".appendVar", ".prependVar", "oe.utils.conditional")
     getvarflags = (".getVarFlag", ".appendVarFlag", ".prependVarFlag")
@@ -212,26 +249,26 @@ class PythonParser():
             funcstr = codegen.to_source(func)
             argstr = codegen.to_source(arg)
         except TypeError:
-            self.log.debug(2, 'Failed to convert function and argument to source form')
+            self.log.debug2('Failed to convert function and argument to source form')
         else:
-            self.log.debug(1, self.unhandled_message % (funcstr, argstr))
+            self.log.debug(self.unhandled_message % (funcstr, argstr))
 
     def visit_Call(self, node):
         name = self.called_node_name(node.func)
         if name and (name.endswith(self.getvars) or name.endswith(self.getvarflags) or name in self.containsfuncs or name in self.containsanyfuncs):
-            if isinstance(node.args[0], ast.Str):
-                varname = node.args[0].s
-                if name in self.containsfuncs and isinstance(node.args[1], ast.Str):
+            if isinstance(node.args[0], ast.Constant) and isinstance(node.args[0].value, str):
+                varname = node.args[0].value
+                if name in self.containsfuncs and isinstance(node.args[1], ast.Constant):
                     if varname not in self.contains:
                         self.contains[varname] = set()
-                    self.contains[varname].add(node.args[1].s)
-                elif name in self.containsanyfuncs and isinstance(node.args[1], ast.Str):
+                    self.contains[varname].add(node.args[1].value)
+                elif name in self.containsanyfuncs and isinstance(node.args[1], ast.Constant):
                     if varname not in self.contains:
                         self.contains[varname] = set()
-                    self.contains[varname].update(node.args[1].s.split())
+                    self.contains[varname].update(node.args[1].value.split())
                 elif name.endswith(self.getvarflags):
-                    if isinstance(node.args[1], ast.Str):
-                        self.references.add('%s[%s]' % (varname, node.args[1].s))
+                    if isinstance(node.args[1], ast.Constant):
+                        self.references.add('%s[%s]' % (varname, node.args[1].value))
                     else:
                         self.warn(node.func, node.args[1])
                 else:
@@ -239,8 +276,8 @@ class PythonParser():
             else:
                 self.warn(node.func, node.args[0])
         elif name and name.endswith(".expand"):
-            if isinstance(node.args[0], ast.Str):
-                value = node.args[0].s
+            if isinstance(node.args[0], ast.Constant):
+                value = node.args[0].value
                 d = bb.data.init()
                 parser = d.expandWithRefs(value, self.name)
                 self.references |= parser.references
@@ -250,8 +287,8 @@ class PythonParser():
                         self.contains[varname] = set()
                     self.contains[varname] |= parser.contains[varname]
         elif name in self.execfuncs:
-            if isinstance(node.args[0], ast.Str):
-                self.var_execs.add(node.args[0].s)
+            if isinstance(node.args[0], ast.Constant):
+                self.var_execs.add(node.args[0].value)
             else:
                 self.warn(node.func, node.args[0])
         elif name and isinstance(node.func, (ast.Name, ast.Attribute)):
@@ -276,16 +313,24 @@ class PythonParser():
         self.contains = {}
         self.execs = set()
         self.references = set()
-        self.log = BufferedLogger('BitBake.Data.PythonParser', logging.DEBUG, log)
+        self._log = log
+        # Defer init as expensive
+        self.log = DummyLogger()
 
         self.unhandled_message = "in call of %s, argument '%s' is not a string literal"
         self.unhandled_message = "while parsing %s, %s" % (name, self.unhandled_message)
 
-    def parse_python(self, node, lineno=0, filename="<string>"):
-        if not node or not node.strip():
+    # For the python module code it is expensive to have the function text so it is
+    # uses a different fixedhash to cache against. We can take the hit on obtaining the
+    # text if it isn't in the cache.
+    def parse_python(self, node, lineno=0, filename="<string>", fixedhash=None):
+        if not fixedhash and (not node or not node.strip()):
             return
 
-        h = bbhash(str(node))
+        if fixedhash:
+            h = fixedhash
+        else:
+            h = bbhash(str(node))
 
         if h in codeparsercache.pythoncache:
             self.references = set(codeparsercache.pythoncache[h].refs)
@@ -303,6 +348,12 @@ class PythonParser():
                 self.contains[i] = set(codeparsercache.pythoncacheextras[h].contains[i])
             return
 
+        if fixedhash and not node:
+            raise KeyError
+
+        # Need to parse so take the hit on the real log buffer
+        self.log = BufferedLogger('BitBake.Data.PythonParser', logging.DEBUG, self._log)
+
         # We can't add to the linenumbers for compile, we can pad to the correct number of blank lines though
         node = "\n" * int(lineno) + node
         code = compile(check_indent(str(node)), filename, "exec",
@@ -321,7 +372,11 @@ class ShellParser():
         self.funcdefs = set()
         self.allexecs = set()
         self.execs = set()
-        self.log = BufferedLogger('BitBake.Data.%s' % name, logging.DEBUG, log)
+        self._name = name
+        self._log = log
+        # Defer init as expensive
+        self.log = DummyLogger()
+
         self.unhandled_template = "unable to handle non-literal command '%s'"
         self.unhandled_template = "while parsing %s, %s" % (name, self.unhandled_template)
 
@@ -340,6 +395,9 @@ class ShellParser():
             self.execs = set(codeparsercache.shellcacheextras[h].execs)
             return self.execs
 
+        # Need to parse so take the hit on the real log buffer
+        self.log = BufferedLogger('BitBake.Data.%s' % self._name, logging.DEBUG, self._log)
+
         self._parse_shell(value)
         self.execs = set(cmd for cmd in self.allexecs if cmd not in self.funcdefs)
 
@@ -450,7 +508,7 @@ class ShellParser():
 
                 cmd = word[1]
                 if cmd.startswith("$"):
-                    self.log.debug(1, self.unhandled_template % cmd)
+                    self.log.debug(self.unhandled_template % cmd)
                 elif cmd == "eval":
                     command = " ".join(word for _, word in words[1:])
                     self._parse_shell(command)