Use CMAKE 3.9 property to enable LTCG instead of manual options.

Link with full warnings, and warnings as errors, on both clang and cl. Enable fast optimizations on clang builds too. Compile with permissive- on VS for standards confirmance Remove #pragmas Remove linker garbage that VS adds Build the DLL as a driver so that it has the correct non-paged pool flags and can be used in KM Remove global variable pointers and leverage global structures directly. This has the side effect of removing the need for any initialized .data section, as well as removing .reloc completely. Remove full PDB name of my computer :) Remove manifest in DLL for kernel compatibility Remove usage of unused unions and structure names. Define __security_check_cookie to make CL happy -- there are no stack buffers at all, and linking with BufferOverflowU.lib simply adds 4KB of UM code that is not called by anything. This also removes the need for an import directory and makes the DLL usable in kernel-mode.
ionescu007 · Apr 18, 2020 · 2df7adf · 2df7adf
1 parent 3341bf2
commit 2df7adf
Show file tree

Hide file tree

Showing 12 changed files with 166 additions and 168 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,4 +1,4 @@
-cmake_minimum_required (VERSION 3.8)
+cmake_minimum_required (VERSION 3.9)
 project (minlzma)
 
 add_subdirectory(minlzlib)

diff --git a/minlzdec/CMakeLists.txt b/minlzdec/CMakeLists.txt
@@ -5,6 +5,10 @@ target_link_libraries(minlzdec LINK_PUBLIC minlzlib)
 
 if(MSVC)
     set(CMAKE_C_STANDARD_LIBRARIES "")
-    set(CMAKE_C_FLAGS_RELWITHDEBINFO "/Ox /Ob2 /Oi /Ot /Oy /GF /GL /Gy /MT /Zi /DNDEBUG")
-    set(CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO "/LTCG /INCREMENTAL:NO /OPT:ICF /OPT:REF /DEBUG")
+    set_property(TARGET minlzdec PROPERTY INTERPROCEDURAL_OPTIMIZATION TRUE)
+    string(REGEX REPLACE "/W[1-3]" "/W4" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}")
+    set(CMAKE_C_FLAGS_RELWITHDEBINFO "/WX /Ox /Ob2 /Oi /Ot /Oy /GF /Gy /MT /Zi /permissive-")
+    set(CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO "/OPT:ICF /OPT:REF /DEBUG /EMITPOGOPHASEINFO /NOVCFEATURE /NOCOFFGRPINFO /PDBALTPATH:minlzdec.pdb")
+else()
+    set(CMAKE_C_FLAGS_RELWITHDEBINFO "-Ofast -Wall -Werror")
 endif()
diff --git a/minlzlib/CMakeLists.txt b/minlzlib/CMakeLists.txt
@@ -1,6 +1,4 @@
-cmake_minimum_required (VERSION 3.8)
-
-set(MINLZLIB_SOURCES "inputbuf.c" "dictbuf.c" "lzma2dec.c" "lzmadec.c" "rangedec.c" "xzstream.c" "lzmadec.h" "xzstream.h" "minlzlib.h")
+set(MINLZLIB_SOURCES "inputbuf.c" "dictbuf.c" "lzma2dec.c" "lzmadec.c" "rangedec.c" "xzstream.c" "lzmadec.h" "xzstream.h" "minlzlib.h")
 add_library(minlz_obj OBJECT ${MINLZLIB_SOURCES})
 set_target_properties(minlz_obj PROPERTIES C_STANDARD 11 C_STANDARD_REQUIRED YES C_EXTENSIONS NO)
 
@@ -13,7 +11,12 @@ target_include_directories (minlz_obj PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
 
 if(MSVC)
     set(CMAKE_C_STANDARD_LIBRARIES "")
-    target_link_libraries(minlz BufferOverflowU kernel32)
-    set(CMAKE_C_FLAGS_RELWITHDEBINFO "/Ox /Ob2 /Oi /Ot /Oy /GF /GL /Gy /MT /Zi")
-    set(CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFO "/LTCG /INCREMENTAL:NO /OPT:ICF /OPT:REF /NODEFAULTLIB /NOENTRY /DEBUG /EXPORT:XzDecode")
+    target_link_libraries(minlz)
+    set_property(TARGET minlz_obj PROPERTY INTERPROCEDURAL_OPTIMIZATION TRUE)
+    set_property(TARGET minlz PROPERTY INTERPROCEDURAL_OPTIMIZATION TRUE)
+    string(REGEX REPLACE "/W[1-3]" "/W4" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}")
+    set(CMAKE_C_FLAGS_RELWITHDEBINFO "/WX /Ox /Ob2 /Oi /Ot /Oy /GF /Gy /MT /Zi /wd4214 /permissive-")
+    set(CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFO "/OPT:ICF /OPT:REF /NODEFAULTLIB /NOENTRY /DEBUG /EXPORT:XzDecode /EMITPOGOPHASEINFO /NOVCFEATURE /NOCOFFGRPINFO /DRIVER /MANIFEST:NO /PDBALTPATH:minlz.pdb /MERGE:.edata=.rdata")
+else()
+    set(CMAKE_C_FLAGS_RELWITHDEBINFO "-Ofast -Wall -Werror")
 endif()
diff --git a/minlzlib/dictbuf.c b/minlzlib/dictbuf.c
@@ -42,8 +42,7 @@ typedef struct _DICTIONARY_STATE
     uint32_t Limit;
     uint32_t Size;
 } DICTIONARY_STATE, *PDICTIONARY_STATE;
-DICTIONARY_STATE g_Dictionary;
-PDICTIONARY_STATE Dictionary = &g_Dictionary;
+DICTIONARY_STATE Dictionary;
 
 void
 DtInitialize (
@@ -54,9 +53,9 @@ DtInitialize (
     //
     // Initialize the buffer and reset the position
     //
-    Dictionary->Buffer = HistoryBuffer;
-    Dictionary->Offset = 0;
-    Dictionary->Size = Size;
+    Dictionary.Buffer = HistoryBuffer;
+    Dictionary.Offset = 0;
+    Dictionary.Size = Size;
 }
 
 bool
@@ -68,12 +67,12 @@ DtSetLimit (
     // Make sure that the passed in dictionary limit fits within the size, and
     // then set this as the new limit. Save the starting point (current offset)
     //
-    if ((Dictionary->Offset + Limit) > Dictionary->Size)
+    if ((Dictionary.Offset + Limit) > Dictionary.Size)
     {
         return false;
     }
-    Dictionary->Limit = Dictionary->Offset + Limit;
-    Dictionary->Start = Dictionary->Offset;
+    Dictionary.Limit = Dictionary.Offset + Limit;
+    Dictionary.Start = Dictionary.Offset;
     return true;
 }
 
@@ -85,8 +84,8 @@ DtIsComplete (
     //
     // Return bytes processed and if the dictionary has been fully written to
     //
-    *BytesProcessed = Dictionary->Offset - Dictionary->Start;
-    return (Dictionary->Offset == Dictionary->Limit);
+    *BytesProcessed = Dictionary.Offset - Dictionary.Start;
+    return (Dictionary.Offset == Dictionary.Limit);
 }
 
 bool
@@ -97,8 +96,8 @@ DtCanWrite (
     //
     // Return our position and make sure it's not beyond the uncompressed size
     //
-    *Position = Dictionary->Offset;
-    return (Dictionary->Offset < Dictionary->Limit);
+    *Position = Dictionary.Offset;
+    return (Dictionary.Offset < Dictionary.Limit);
 }
 
 uint8_t
@@ -110,11 +109,11 @@ DtGetSymbol (
     // If the dictionary is still empty, just return 0, otherwise, return the
     // symbol that is Distance bytes backward
     //
-    if (Distance > Dictionary->Offset)
+    if (Distance > Dictionary.Offset)
     {
         return 0;
     }
-    return Dictionary->Buffer[Dictionary->Offset - Distance];
+    return Dictionary.Buffer[Dictionary.Offset - Distance];
 }
 
 void
@@ -125,7 +124,7 @@ DtPutSymbol (
     //
     // Write the symbol and advance our position
     //
-    Dictionary->Buffer[Dictionary->Offset++] = Symbol;
+    Dictionary.Buffer[Dictionary.Offset++] = Symbol;
 }
 
 bool
@@ -138,7 +137,7 @@ DtRepeatSymbol (
     // Make sure we never get asked to write past the end of the dictionary,
     // then rewrite the stream of symbols forward into the dictionary
     //
-    if ((Length + Dictionary->Offset) > Dictionary->Limit)
+    if ((Length + Dictionary.Offset) > Dictionary.Limit)
     {
         return false;
     }

diff --git a/minlzlib/inputbuf.c b/minlzlib/inputbuf.c
@@ -39,8 +39,7 @@ typedef struct _BUFFER_STATE
     uint32_t Offset;
     uint32_t Size;
 } BUFFER_STATE, * PBUFFER_STATE;
-BUFFER_STATE g_In;
-PBUFFER_STATE In = &g_In;
+BUFFER_STATE In;
 
 bool
 BfAlign (
@@ -51,7 +50,7 @@ BfAlign (
     //
     // Keep reading until we reach 32-bit alignment. All bytes must be zero.
     //
-    while (In->Offset & 3)
+    while (In.Offset & 3)
     {
         if (!BfRead(&padByte) || (padByte != 0))
         {
@@ -71,13 +70,13 @@ BfSeek (
     // Make sure the input buffer has enough space to seek the desired size, if
     // it does, return the current position and then seek past the desired size
     //
-    if ((In->Offset + Length) > In->Size)
+    if ((In.Offset + Length) > In.Size)
     {
         *Bytes = 0;
         return false;
     }
-    *Bytes = &In->Buffer[In->Offset];
-    In->Offset += Length;
+    *Bytes = &In.Buffer[In.Offset];
+    In.Offset += Length;
     return true;
 }
 
@@ -108,7 +107,7 @@ BfInitialize (
     //
     // Save all the data in the context buffer state
     //
-    In->Buffer = InputBuffer;
-    In->Size = InputSize;
-    In->Offset = 0;
+    In.Buffer = InputBuffer;
+    In.Size = InputSize;
+    In.Offset = 0;
 }
diff --git a/minlzlib/lzma2dec.c b/minlzlib/lzma2dec.c
@@ -39,8 +39,7 @@ typedef struct _CHUNK_STATE
     uint32_t UncompressedSize;
     uint16_t CompressedSize;
 } CHUNK_STATE, * PCHUNK_STATE;
-CHUNK_STATE g_ChunkState;
-PCHUNK_STATE ChunkState = &g_ChunkState;
+CHUNK_STATE ChunkState;
 
 bool
 Lz2DecodeChunk (
@@ -51,7 +50,7 @@ Lz2DecodeChunk (
     //
     // Make sure we always have space for the biggest possible LZMA sequence
     //
-    if (ChunkState->CompressedSize < LZMA_MAX_SEQUENCE_SIZE)
+    if (ChunkState.CompressedSize < LZMA_MAX_SEQUENCE_SIZE)
     {
         return false;
     }
@@ -70,7 +69,7 @@ Lz2DecodeChunk (
     // exactly where we expected it to.
     //
     if (!RcIsComplete(&bytesProcessed) ||
-        (bytesProcessed != ChunkState->CompressedSize))
+        (bytesProcessed != ChunkState.CompressedSize))
     {
         return false;
     }
@@ -80,7 +79,7 @@ Lz2DecodeChunk (
     // must be full now.
     //
     if (!DtIsComplete(&bytesProcessed) ||
-        (bytesProcessed != ChunkState->UncompressedSize))
+        (bytesProcessed != ChunkState.UncompressedSize))
     {
         return false;
     }
@@ -115,7 +114,7 @@ Lz2DecodeStream (
         //
         // We only support compressed streams, not uncompressed ones
         //
-        if (controlByte.Common.IsLzma == 0)
+        if (controlByte.u.Common.IsLzma == 0)
         {
             break;
         }
@@ -132,10 +131,10 @@ Lz2DecodeStream (
         // Decode the 1-based big-endian uncompressed size. It must fit into
         // the output buffer that was supplied.
         //
-        ChunkState->UncompressedSize = controlByte.Lzma.UncompressedSize << 16;
-        ChunkState->UncompressedSize += (*pInfoBytes)[0] << 8;
-        ChunkState->UncompressedSize += (*pInfoBytes)[1] + 1;
-        if (!DtSetLimit(ChunkState->UncompressedSize))
+        ChunkState.UncompressedSize = controlByte.u.Lzma.UncompressedSize << 16;
+        ChunkState.UncompressedSize += (*pInfoBytes)[0] << 8;
+        ChunkState.UncompressedSize += (*pInfoBytes)[1] + 1;
+        if (!DtSetLimit(ChunkState.UncompressedSize))
         {
             break;
         }
@@ -144,14 +143,14 @@ Lz2DecodeStream (
         // Decode the 1-based big-endian compressed size. It must fit into the
         // input buffer that was given, which RcInitialize will verify below.
         //
-        ChunkState->CompressedSize = (*pInfoBytes)[2] << 8;
-        ChunkState->CompressedSize += (*pInfoBytes)[3] + 1;
+        ChunkState.CompressedSize = (*pInfoBytes)[2] << 8;
+        ChunkState.CompressedSize += (*pInfoBytes)[3] + 1;
 
         //
         // Check if the full LZMA state needs to be reset, which must happen at
         // the start of stream. Other reset states are not supported.
         //
-        if (controlByte.Lzma.ResetState == Lzma2FullReset)
+        if (controlByte.u.Lzma.ResetState == Lzma2FullReset)
         {
             //
             // Read the LZMA properties and then initialize the decoder.
@@ -161,7 +160,7 @@ Lz2DecodeStream (
                 break;
             }
         }
-        else if (controlByte.Lzma.ResetState != Lzma2NoReset)
+        else if (controlByte.u.Lzma.ResetState != Lzma2NoReset)
         {
             break;
         }
@@ -171,7 +170,7 @@ Lz2DecodeStream (
         // coding decoder, and let it know how much input data exists. We've
         // already validated that this much space exists in the input buffer.
         //
-        if (!RcInitialize(&ChunkState->CompressedSize))
+        if (!RcInitialize(&ChunkState.CompressedSize))
         {
             break;
         }

diff --git a/minlzlib/lzma2dec.h b/minlzlib/lzma2dec.h
@@ -64,9 +64,6 @@ typedef enum _LZMA2_COMPRESSED_RESET_STATE
 //
 // This describes how an LZMA2 control byte can be parsed
 //
-#pragma warning(push)
-#pragma warning(disable:4201)
-#pragma warning(disable:4214)
 typedef union _LZMA2_CONTROL_BYTE
 {
     union
@@ -90,8 +87,7 @@ typedef union _LZMA2_CONTROL_BYTE
             uint8_t ResetState : 2;
             uint8_t IsLzma : 1;
         } Common;
-    };
+    } u;
     uint8_t Value;
 } LZMA2_CONTROL_BYTE;
 static_assert(sizeof(LZMA2_CONTROL_BYTE) == 1, "Invalid control byte size");
-#pragma warning(pop)