audk/BaseTools/Source/Python/Workspace/WorkspaceCommon.py
Liming Gao 3a4698202c Revert "BaseTools:BaseTools supports to the driver combination."
This reverts commit 838bc257bae3f9fc6723f41f3980f6cfbedb77e5.
After further evaluation, there are the unclear behavior in for the
driver combination feature. To not impact Q1 stable tag, remove it first.
1. If the drivers to be combined have the different PCD or library instance
   setting, build should not combine them and report build break. But this
   commit doesn't consider this case.
2. When start the sub driver fail, continue to start other sub driver. This
   behavior is required to be clarifed in build spec.
3. Unload the sub driver when the combined driver start fail. This case need
   to call the sub driver unload function for the driver start fail only.

Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Liming Gao <liming.gao@intel.com>
Reviewed-by: Bob Feng <bob.c.feng@intel.com>
2019-02-28 10:01:53 +08:00

255 lines
11 KiB
Python

## @file
# Common routines used by workspace
#
# Copyright (c) 2012 - 2018, Intel Corporation. All rights reserved.<BR>
# This program and the accompanying materials
# are licensed and made available under the terms and conditions of the BSD License
# which accompanies this distribution. The full text of the license may be found at
# http://opensource.org/licenses/bsd-license.php
#
# THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
# WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
#
from __future__ import absolute_import
from collections import OrderedDict, defaultdict
from Common.DataType import SUP_MODULE_USER_DEFINED
from .BuildClassObject import LibraryClassObject
import Common.GlobalData as GlobalData
from Workspace.BuildClassObject import StructurePcd
from Common.BuildToolError import RESOURCE_NOT_AVAILABLE
from Common.BuildToolError import OPTION_MISSING
from Common.BuildToolError import BUILD_ERROR
class OrderedListDict(OrderedDict):
def __init__(self, *args, **kwargs):
super(OrderedListDict, self).__init__(*args, **kwargs)
self.default_factory = list
def __missing__(self, key):
self[key] = Value = self.default_factory()
return Value
## Get all packages from platform for specified arch, target and toolchain
#
# @param Platform: DscBuildData instance
# @param BuildDatabase: The database saves all data for all metafiles
# @param Arch: Current arch
# @param Target: Current target
# @param Toolchain: Current toolchain
# @retval: List of packages which are DecBuildData instances
#
def GetPackageList(Platform, BuildDatabase, Arch, Target, Toolchain):
PkgSet = set()
for ModuleFile in Platform.Modules:
Data = BuildDatabase[ModuleFile, Arch, Target, Toolchain]
PkgSet.update(Data.Packages)
for Lib in GetLiabraryInstances(Data, Platform, BuildDatabase, Arch, Target, Toolchain):
PkgSet.update(Lib.Packages)
return list(PkgSet)
## Get all declared PCD from platform for specified arch, target and toolchain
#
# @param Platform: DscBuildData instance
# @param BuildDatabase: The database saves all data for all metafiles
# @param Arch: Current arch
# @param Target: Current target
# @param Toolchain: Current toolchain
# @retval: A dictionary contains instances of PcdClassObject with key (PcdCName, TokenSpaceGuid)
# @retval: A dictionary contains real GUIDs of TokenSpaceGuid
#
def GetDeclaredPcd(Platform, BuildDatabase, Arch, Target, Toolchain, additionalPkgs):
PkgList = GetPackageList(Platform, BuildDatabase, Arch, Target, Toolchain)
PkgList = set(PkgList)
PkgList |= additionalPkgs
DecPcds = {}
GuidDict = {}
for Pkg in PkgList:
Guids = Pkg.Guids
GuidDict.update(Guids)
for Pcd in Pkg.Pcds:
PcdCName = Pcd[0]
PcdTokenName = Pcd[1]
if GlobalData.MixedPcd:
for PcdItem in GlobalData.MixedPcd:
if (PcdCName, PcdTokenName) in GlobalData.MixedPcd[PcdItem]:
PcdCName = PcdItem[0]
break
if (PcdCName, PcdTokenName) not in DecPcds:
DecPcds[PcdCName, PcdTokenName] = Pkg.Pcds[Pcd]
return DecPcds, GuidDict
## Get all dependent libraries for a module
#
# @param Module: InfBuildData instance
# @param Platform: DscBuildData instance
# @param BuildDatabase: The database saves all data for all metafiles
# @param Arch: Current arch
# @param Target: Current target
# @param Toolchain: Current toolchain
# @retval: List of dependent libraries which are InfBuildData instances
#
def GetLiabraryInstances(Module, Platform, BuildDatabase, Arch, Target, Toolchain):
return GetModuleLibInstances(Module, Platform, BuildDatabase, Arch, Target, Toolchain)
def GetModuleLibInstances(Module, Platform, BuildDatabase, Arch, Target, Toolchain, FileName = '', EdkLogger = None):
ModuleType = Module.ModuleType
# add forced library instances (specified under LibraryClasses sections)
#
# If a module has a MODULE_TYPE of USER_DEFINED,
# do not link in NULL library class instances from the global [LibraryClasses.*] sections.
#
if Module.ModuleType != SUP_MODULE_USER_DEFINED:
for LibraryClass in Platform.LibraryClasses.GetKeys():
if LibraryClass.startswith("NULL") and Platform.LibraryClasses[LibraryClass, Module.ModuleType]:
Module.LibraryClasses[LibraryClass] = Platform.LibraryClasses[LibraryClass, Module.ModuleType]
# add forced library instances (specified in module overrides)
for LibraryClass in Platform.Modules[str(Module)].LibraryClasses:
if LibraryClass.startswith("NULL"):
Module.LibraryClasses[LibraryClass] = Platform.Modules[str(Module)].LibraryClasses[LibraryClass]
# EdkII module
LibraryConsumerList = [Module]
Constructor = []
ConsumedByList = OrderedListDict()
LibraryInstance = OrderedDict()
if FileName:
EdkLogger.verbose("")
EdkLogger.verbose("Library instances of module [%s] [%s]:" % (str(Module), Arch))
while len(LibraryConsumerList) > 0:
M = LibraryConsumerList.pop()
for LibraryClassName in M.LibraryClasses:
if LibraryClassName not in LibraryInstance:
# override library instance for this module
LibraryPath = Platform.Modules[str(Module)].LibraryClasses.get(LibraryClassName,Platform.LibraryClasses[LibraryClassName, ModuleType])
if LibraryPath is None:
LibraryPath = M.LibraryClasses.get(LibraryClassName)
if LibraryPath is None:
if FileName:
EdkLogger.error("build", RESOURCE_NOT_AVAILABLE,
"Instance of library class [%s] is not found" % LibraryClassName,
File=FileName,
ExtraData="in [%s] [%s]\n\tconsumed by module [%s]" % (str(M), Arch, str(Module)))
else:
return []
LibraryModule = BuildDatabase[LibraryPath, Arch, Target, Toolchain]
# for those forced library instance (NULL library), add a fake library class
if LibraryClassName.startswith("NULL"):
LibraryModule.LibraryClass.append(LibraryClassObject(LibraryClassName, [ModuleType]))
elif LibraryModule.LibraryClass is None \
or len(LibraryModule.LibraryClass) == 0 \
or (ModuleType != SUP_MODULE_USER_DEFINED
and ModuleType not in LibraryModule.LibraryClass[0].SupModList):
# only USER_DEFINED can link against any library instance despite of its SupModList
if FileName:
EdkLogger.error("build", OPTION_MISSING,
"Module type [%s] is not supported by library instance [%s]" \
% (ModuleType, LibraryPath), File=FileName,
ExtraData="consumed by [%s]" % str(Module))
else:
return []
LibraryInstance[LibraryClassName] = LibraryModule
LibraryConsumerList.append(LibraryModule)
if FileName:
EdkLogger.verbose("\t" + str(LibraryClassName) + " : " + str(LibraryModule))
else:
LibraryModule = LibraryInstance[LibraryClassName]
if LibraryModule is None:
continue
if LibraryModule.ConstructorList != [] and LibraryModule not in Constructor:
Constructor.append(LibraryModule)
# don't add current module itself to consumer list
if M != Module:
if M in ConsumedByList[LibraryModule]:
continue
ConsumedByList[LibraryModule].append(M)
#
# Initialize the sorted output list to the empty set
#
SortedLibraryList = []
#
# Q <- Set of all nodes with no incoming edges
#
LibraryList = [] #LibraryInstance.values()
Q = []
for LibraryClassName in LibraryInstance:
M = LibraryInstance[LibraryClassName]
LibraryList.append(M)
if not ConsumedByList[M]:
Q.append(M)
#
# start the DAG algorithm
#
while True:
EdgeRemoved = True
while Q == [] and EdgeRemoved:
EdgeRemoved = False
# for each node Item with a Constructor
for Item in LibraryList:
if Item not in Constructor:
continue
# for each Node without a constructor with an edge e from Item to Node
for Node in ConsumedByList[Item]:
if Node in Constructor:
continue
# remove edge e from the graph if Node has no constructor
ConsumedByList[Item].remove(Node)
EdgeRemoved = True
if not ConsumedByList[Item]:
# insert Item into Q
Q.insert(0, Item)
break
if Q != []:
break
# DAG is done if there's no more incoming edge for all nodes
if Q == []:
break
# remove node from Q
Node = Q.pop()
# output Node
SortedLibraryList.append(Node)
# for each node Item with an edge e from Node to Item do
for Item in LibraryList:
if Node not in ConsumedByList[Item]:
continue
# remove edge e from the graph
ConsumedByList[Item].remove(Node)
if ConsumedByList[Item]:
continue
# insert Item into Q, if Item has no other incoming edges
Q.insert(0, Item)
#
# if any remaining node Item in the graph has a constructor and an incoming edge, then the graph has a cycle
#
for Item in LibraryList:
if ConsumedByList[Item] and Item in Constructor and len(Constructor) > 1:
if FileName:
ErrorMessage = "\tconsumed by " + "\n\tconsumed by ".join(str(L) for L in ConsumedByList[Item])
EdkLogger.error("build", BUILD_ERROR, 'Library [%s] with constructors has a cycle' % str(Item),
ExtraData=ErrorMessage, File=FileName)
else:
return []
if Item not in SortedLibraryList:
SortedLibraryList.append(Item)
#
# Build the list of constructor and destructor names
# The DAG Topo sort produces the destructor order, so the list of constructors must generated in the reverse order
#
SortedLibraryList.reverse()
return SortedLibraryList