CHANGES

Minor fixes for C++11 and use of std::atomic
Travis: change to precise linux, we don't trust trusty
2017-03-01 18:24:01 -08:00 · 2016-11-07 22:42:11 -08:00 · 2016-11-07 22:42:11 -08:00 · 2016-10-31 15:48:19 -07:00 · 2016-10-31 15:48:19 -07:00 · 2016-10-31 15:48:19 -07:00
87 changed files with 717 additions and 442 deletions
@@ -8,21 +8,21 @@ compiler:
 os:
    - linux
    - osx
-#osx_image: xcode7
-dist: trusty
+osx_image: xcode6.4
+dist: precise

 # Do a default build (C++03, sse2), an sse4.2 build, and a C++11 build.
 # Also try with both gcc 4.8 and 4.9 (on Linux; on OSX these will be
 # excluded).
 env:
    matrix:
-      - WHICHGCC=4.8 BUILD_FLAGS=""
-      - WHICHGCC=4.8 BUILD_FLAGS="USE_CPP11=1"
-      - WHICHGCC=4.9 BUILD_FLAGS=""
-      - WHICHGCC=5 BUILD_FLAGS="USE_CPP11=1"
-      - WHICHGCC=4.8 OIIOBRANCH=release BUILD_FLAGS=""
-      - WHICHGCC=4.8 OIIOBRANCH=release BUILD_FLAGS="USE_CPP11=1"
-      - WHICHGCC=4.8 BUILD_FLAGS="" DEBUG=1
+      - WHICHGCC=4.8 USE_CPP11=1
+      - WHICHGCC=4.9 USE_CPP11=1
+      - WHICHGCC=5 USE_CPP11=1
+      - WHICHGCC=6 USE_CPP11=1
+      - WHICHGCC=4.8 OIIOBRANCH=RB-1.7 USE_CPP11=0
+      - WHICHGCC=4.8 OIIOBRANCH=RB-1.7 USE_CPP11=1
+      - WHICHGCC=4.8 DEBUG=1 USE_CPP11=1

 # Add-ons: specify apt packages for Linux
 addons:
@@ -37,6 +37,8 @@ addons:
      - g++-4.9
      - gcc-5
      - g++-5
+      - gcc-6
+      - g++-6
      - libboost1.55-all-dev
      - libtiff4-dev
      - llvm3.4
@@ -56,6 +58,7 @@ before_install:
          export PLATFORM=linux64 ;
          cat /proc/cpuinfo ;
      fi
+    - export OIIOPLATFORM=$PLATFORM
    - if [ $DEBUG == 1 ] ; then export PLATFORM=${PLATFORM}.debug ; fi
    - echo "Build platform name is $PLATFORM"

@@ -71,9 +74,9 @@ install:
          export ILMBASE_HOME=$PWD/openexr-install ;
          export OPENEXR_HOME=$PWD/openexr-install ;
      fi
-    - export OIIOMAKEFLAGS="$OIIOMAKEFLAGS -j2"
+    - export OIIOMAKEFLAGS="$OIIOMAKEFLAGS -j2 DEBUG= "
    - src/build-scripts/build_openimageio.bash
-    - export OPENIMAGEIOHOME=$PWD/OpenImageIO/dist/$PLATFORM
+    - export OPENIMAGEIOHOME=$PWD/OpenImageIO/dist/$OIIOPLATFORM
    - export DYLD_LIBRARY_PATH=$OPENIMAGEIOHOME/lib:$DYLD_LIBRARY_PATH
    - export LD_LIBRARY_PATH=$OPENIMAGEIOHOME/lib:$LD_LIBRARY_PATH
    - export PATH=$OPENIMAGEIOHOME/bin:$PATH
@@ -113,10 +116,13 @@ matrix:
        compiler: clang
      - os: osx
        compiler: clang
-        env: WHICHGCC=4.9 BUILD_FLAGS=""
+        env: WHICHGCC=4.9 USE_CPP11=1
      - os: osx
        compiler: clang
-        env: WHICHGCC=5 BUILD_FLAGS="USE_CPP11=1"
+        env: WHICHGCC=5 USE_CPP11=1
+      - os: osx
+        compiler: clang
+        env: WHICHGCC=6 USE_CPP11=1

 notifications:
    email:
@@ -1,6 +1,59 @@
 Changes:

-Release 1.7 -- 29 Jan 2016 (compared to 1.6)
+Release 1.7.6 -- 1 Mar 2017(compared to 1.7.5)
+--------------------------------------------------
+* Minor fixes for C++11 and use of std::atomic.
+
+Release 1.7.5 -- 1 Nov 2016 (compared to 1.7.4)
+--------------------------------------------------
+* Fix oslc crash/assertion when indexing into a closure type. #663
+* A version of spline() where the knots were colors without derivs
+  and the abscissa was a float with derivs, computed its result
+  derivatives incorrectly. #666
+* Incorrect dropping of certain optional texture arguments. #671
+* Bug fix: matrix parameters initialized with matrix(space,1) did not
+  initialize properly. #675
+* Suppress warnings when compiling with clang 3.9. #683
+* oslc: better error reporting for use of non-function symbol as a
+  function. #684
+* Filed to properly propagate the C++ standard flags (e.g. -std=c++11)
+  when turning llvm_ops.cpp into LLVM bitcode. #674
+
+Release 1.7.4 -- 1 Aug 2016 (compared to 1.7.3)
+--------------------------------------------------
+* Bug fix: incorrect instance merging when unspecified-length arrays
+  differed. (#656)
+* Make oslc understand hex integer constants (like 0x01fc). #653
+
+Release 1.7.3 -- 1 Jul 2016 (compared to 1.7.2)
+--------------------------------------------------
+* ShadingSystem attribute "userdata_isconnected", if set to a nonzero
+  integer, causes a shader call to isconnected(param) to return 1 if the
+  parameter is potentially connected to userdata (that is, if lockgeom=0
+  for that parameter). This if for particular renderer needs, and the
+  default remains for isconnected() to return 1 only if there is an
+  actual connection. #629 (1.8.0/1.7.3)
+* oslc now will warn when assigning the result of a "comma operator",
+  which almost always is a programming error where the shader writer
+  omitted the type name when attempting to construct a vector or matrix.
+  #635 (1.8.0/1.7.3)
+* Fix oslc incorrectly emitting init ops for constructor color(float).
+  #637 (1.8.0/1.7.3)
+* Bug fix: don't test or set the "layer ran bits" on the last layer of the
+  group. #626 (1.8.0/1.7.3)
+* Windows fixes: build breaks #643; fixes for lexer #646.
+* Fixes for clean compile with gcc 6.x. #647
+* Fix subtle bug in runtime optimizer that could seg fault. #651
+
+Release 1.7.2 -- 1 Mar 2016 (compared to 1.7.1)
+--------------------------------------------------
+* Build: better handling of multiple USE_SIMD options. #617
+* oslc bug: getmatrix() failed to note that the matrix argument was
+  written. #618
+* CMake FindOpenEXR.cmake totally rewritten, much simpler and better. #621
+
+
+Release 1.7 [1.7.1] -- 29 Jan 2016 (compared to 1.6)
 -----------------------------------------------
 Language, standard libary, and compiler changes (for shader writers):
 * Language features:
@@ -1327,6 +1380,3 @@ Release 0.5.2 - Mar 14, 2011
  "smartcubic", "cubic", "linear", "closest").
 * Bug fixes to getmessage() and its handling of derivatives, which includes
  a slight RendererServices API change.
-
-
-
@@ -3,8 +3,8 @@ project (OSL)
 # Release version of the library
 set (OSL_LIBRARY_VERSION_MAJOR 1)
 set (OSL_LIBRARY_VERSION_MINOR 7)
-set (OSL_LIBRARY_VERSION_PATCH 1)
-set (OSL_LIBRARY_VERSION_RELEASE_TYPE "dev")   # "dev", "betaX", "RCY", ""
+set (OSL_LIBRARY_VERSION_PATCH 6)
+set (OSL_LIBRARY_VERSION_RELEASE_TYPE "")   # "dev", "betaX", "RCY", ""

 # Version of the OSO file format and instruction set
 set (OSO_FILE_VERSION_MAJOR 1)
@@ -152,6 +152,10 @@ if (CMAKE_COMPILER_IS_CLANG OR CMAKE_COMPILER_IS_APPLECLANG)
        APPLECLANG_VERSION_STRING VERSION_GREATER 6.1)
        add_definitions ("-Wno-unused-local-typedefs")
    endif ()
+    if (CLANG_VERSION_STRING VERSION_EQUAL 3.9 OR CLANG_VERSION_STRING VERSION_GREATER 3.9)
+        # Don't warn about using unknown preprocessor symbols in #if'set
+        add_definitions ("-Wno-expansion-to-defined")
+    endif ()
 endif ()

 # gcc specific options
@@ -260,6 +264,11 @@ else ()
 	add_definitions ("-DOSL_FAST_MATH=0")
 endif ()

+set (USE_OIIO_STATIC ON CACHE BOOL "If OIIO is built static")
+if (USE_OIIO_STATIC)
+    add_definitions ("-DOIIO_STATIC_BUILD=1")
+endif ()
+
 if (OSL_NO_DEFAULT_TEXTURESYSTEM)
    add_definitions ("-DOSL_NO_DEFAULT_TEXTURESYSTEM=1")
 endif ()
@@ -273,10 +282,12 @@ if (OSL_BUILD_CPP11 OR OSL_BUILD_CPP14)
    if (OSL_BUILD_CPP11)
        message (STATUS "Building for C++11")
        add_definitions ("-std=c++11")
+        set (OSL_CSTD_FLAGS ${OSL_CSTD_FLAGS} "-std=c++11")
    endif()
    if (OSL_BUILD_CPP14)
        message (STATUS "Building for C++14")
        add_definitions ("-std=c++14")
+        set (OSL_CSTD_FLAGS ${OSL_CSTD_FLAGS} "-std=c++14")
    endif()
    if (CMAKE_COMPILER_IS_CLANG)
        # C++ >= 11 doesn't like 'register' keyword, which is in Qt headers
@@ -289,10 +300,12 @@ if (USE_LIBCPLUSPLUS AND CMAKE_COMPILER_IS_CLANG)
    add_definitions ("-stdlib=libc++")
 endif ()

+set (OSL_SIMD_FLAGS "")
 if (NOT USE_SIMD STREQUAL "")
    message (STATUS "Compiling with SIMD level ${USE_SIMD}")
    if (USE_SIMD STREQUAL "0")
        add_definitions ("-DOIIO_NO_SSE=1")
+        set (OSL_SIMD_FLAGS ${OSL_SIMD_FLAGS} "-DOIIO_NO_SSE=1")
    else ()
        string (REPLACE "," ";" SIMD_FEATURE_LIST ${USE_SIMD})
        foreach (feature ${SIMD_FEATURE_LIST})
@@ -301,8 +314,10 @@ if (NOT USE_SIMD STREQUAL "")
            endif ()
            if (MSVC OR CMAKE_COMPILER_IS_INTEL)
                add_definitions ("/arch:${feature}")
+                set (OSL_SIMD_FLAGS ${OSL_SIMD_FLAGS} "/arch:${feature}")
            else ()
                add_definitions ("-m${feature}")
+                set (OSL_SIMD_FLAGS ${OSL_SIMD_FLAGS} "-m${feature}")
            endif ()
        endforeach()
    endif ()
@@ -516,11 +531,13 @@ TESTSUITE ( and-or-not-synonyms aastep arithmetic array array-derivs array-range
            function-earlyreturn function-simple function-outputelem
            geomath getattribute-camera getsymbol-nonheap gettextureinfo
            group-outputs groupstring
-            hyperb ieee_fp if incdec initops intbits isconnected isconstant
+            hex hyperb
+            ieee_fp if incdec initops intbits isconnected isconstant
            layers layers-Ciassign layers-entry layers-lazy
            layers-nonlazycopy layers-repeatedoutputs
            linearstep
-            logic loop matrix message mergeinstances-nouserdata
+            logic loop matrix message
+            mergeinstances-nouserdata mergeinstances-vararray
            metadata-braces miscmath missing-shader
            noise noise-cell
            noise-gabor noise-gabor2d-filter noise-gabor3d-filter
@@ -529,24 +546,28 @@ TESTSUITE ( and-or-not-synonyms aastep arithmetic array array-derivs array-range
            oslc-comma oslc-D
            oslc-err-arrayindex oslc-err-closuremul
            oslc-err-format oslc-err-intoverflow
-            oslc-err-noreturn oslc-err-outputparamvararray oslc-err-paramdefault
+            oslc-err-noreturn oslc-err-notfunc
+            oslc-err-outputparamvararray oslc-err-paramdefault
            oslc-err-struct-array-init oslc-err-struct-dup
+            oslc-warn-commainit
            oslc-variadic-macro
-            oslinfo-arrayparams oslinfo-metadata oslinfo-noparams
+            oslinfo-arrayparams oslinfo-colorctrfloat
+            oslinfo-metadata oslinfo-noparams
            osl-imageio
            printf-whole-array
            raytype reparam
            render-background render-bumptest
            render-cornell render-furnace-diffuse
            render-microfacet render-oren-nayar render-veachmis render-ward
-            shortcircuit spline splineinverse string 
+            shortcircuit spline splineinverse spline-derivbug
+            string
            struct struct-array struct-array-mixture
            struct-err struct-in-struct-init struct-init-copy struct-layers
            struct-with-array struct-within-struct
            ternary
            testshade-expr
-            texture-alpha texture-blur texture-derivs
-            texture-firstchannel texture-interp 
+            texture-alpha texture-blur texture-connected-options
+            texture-derivs texture-firstchannel texture-interp
            texture-missingcolor texture-simple
            texture-smallderivs texture-swirl
            texture-width texture-withderivs texture-wrap
@@ -98,6 +98,10 @@ ifneq (${USE_FAST_MATH},)
 MY_CMAKE_FLAGS += -DUSE_FAST_MATH:BOOL=${USE_FAST_MATH}
 endif

+ifneq (${OPENEXR_HOME},)
+MY_CMAKE_FLAGS += -DOPENEXR_HOME:STRING=${OPENEXR_HOME}
+endif
+
 ifneq (${ILMBASE_HOME},)
 MY_CMAKE_FLAGS += -DILMBASE_HOME:STRING=${ILMBASE_HOME}
 endif
@@ -328,6 +332,7 @@ help:
 	@echo "      LINKSTATIC=1             Link with static external libraries when possible"
 	@echo "  Finding and Using Dependencies:"
 	@echo "      BOOST_HOME=path          Custom Boost installation"
+	@echo "      OPENEXR_HOME=path        Custom OpenEXR installation"
 	@echo "      ILMBASE_HOME=path        Custom Ilmbase installation"
 	@echo "      PARTIO_HOME=             Use Partio from the given location"
 	@echo "      USE_EXTERNAL_PUGIXML=1   Use the system PugiXML, not the one in OIIO"
@@ -11,9 +11,8 @@ MY_CMAKE_FLAGS += -DHIDE_SYMBOLS:BOOL=1
 MY_CMAKE_FLAGS += -DBUILDSTATIC:BOOL=1
 MY_CMAKE_FLAGS += -DLINKSTATIC:BOOL=1
 MY_CMAKE_FLAGS += -DUSE_FAST_MATH:BOOL=1
-MY_CMAKE_FLAGS += -DILMBASE_LIB_AREA:STRING=${ILMBASE_HOME}/lib/linux-ix86-64/gcc4.1
-MY_CMAKE_FLAGS += -DILMBASE_CUSTOM:BOOL=True
-MY_CMAKE_FLAGS += -DILMBASE_CUSTOM_LIBRARIES:STRING="Half-fPIC Iex-fPIC IexMath-fPIC IlmThread-fPIC  Imath-fPIC"
+MY_CMAKE_FLAGS += -DOPENEXR_CUSTOM_LIB_DIR:STRING=${ILMBASE_HOME}/lib/linux-ix86-64/gcc4.1
+MY_CMAKE_FLAGS += -DOPENEXR_CUSTOM_LIB_SUFFIX:STRING="-fPIC"

 # Don't build testshade or testrender (mostly because they require the
 # image-handling parts of OIIO, which Pixar doesn't build)
@@ -7,6 +7,7 @@ MY_CMAKE_FLAGS += -DEXTRA_CPP_ARGS="-DOSL_SPI=1"

 ## Generic OSX machines (including LG's laptop)
 ifeq (${platform}, macosx)
+    USE_CPP11 ?= 1
    MY_CMAKE_FLAGS += \
        -DCMAKE_BUILD_WITH_INSTALL_RPATH=1 \
        -DCMAKE_INSTALL_NAME_DIR="${working_dir}/dist/${platform}${variant}/lib"
@@ -86,24 +87,13 @@ ifeq ($(SP_OS), spinux1)
    # Our minimal architecture for workstations and farm supports SSE 4.1
    USE_SIMD = sse4.1

-    ifneq (,$(wildcard /usr/include/OpenEXR2))
    MY_CMAKE_FLAGS += \
-	  -DILMBASE_CUSTOM=1 \
-	  -DILMBASE_CUSTOM_INCLUDE_DIR=/usr/include/OpenEXR2 \
-          -DILMBASE_CUSTOM_LIB_DIR=/usr/lib64/OpenEXR2 \
-	  -DILMBASE_CUSTOM_LIBRARIES="Imath Half IlmThread Iex"
-    else
-        MY_CMAKE_FLAGS += \
-	  -DILMBASE_CUSTOM=1 \
-          -DILMBASE_CUSTOM_LIB_DIR=/usr/lib64 \
-	  -DILMBASE_CUSTOM_LIBRARIES="SpiImath SpiHalf SpiIlmThread SpiIex"
-    endif
-
-    MY_CMAKE_FLAGS += \
-        -DLLVM_CUSTOM=1 \
-	-DLLVM_STATIC=1 \
+	-DOPENEXR_CUSTOM_INCLUDE_DIR=/usr/include/OpenEXR2 \
+	-DOPENEXR_CUSTOM_LIB_DIR=/usr/lib64/OpenEXR2 \
 	-DEXTRA_OSLEXEC_LIBRARIES="/usr/lib64/libpthread.so" \
 	-DEXTRA_OSLEXEC_LIBRARIES="/usr/lib64/libcurses.so" \
+        -DLLVM_CUSTOM=1 \
+	-DLLVM_STATIC=1 \
        -DLLVM_VERSION=${LLVM_VERSION} \
        -DLLVM_DIRECTORY=${LLVM_DIRECTORY}

@@ -172,24 +162,13 @@ ifeq ($(SP_OS), rhel7)
    # Our minimal architecture for workstations and farm supports SSE 4.1
    USE_SIMD = sse4.1

-    ifneq (,$(wildcard /usr/include/OpenEXR2))
    MY_CMAKE_FLAGS += \
-	  -DILMBASE_CUSTOM=1 \
-	  -DILMBASE_CUSTOM_INCLUDE_DIR=/usr/include/OpenEXR2 \
-          -DILMBASE_CUSTOM_LIB_DIR=/usr/lib64/OpenEXR2 \
-	  -DILMBASE_CUSTOM_LIBRARIES="Imath Half IlmThread Iex"
-    else
-        MY_CMAKE_FLAGS += \
-	  -DILMBASE_CUSTOM=1 \
-          -DILMBASE_CUSTOM_LIB_DIR=/usr/lib64 \
-	  -DILMBASE_CUSTOM_LIBRARIES="SpiImath SpiHalf SpiIlmThread SpiIex"
-    endif
-
-    MY_CMAKE_FLAGS += \
-        -DLLVM_CUSTOM=1 \
-	-DLLVM_STATIC=1 \
+	-DOPENEXR_CUSTOM_INCLUDE_DIR=/usr/include/OpenEXR2 \
+	-DOPENEXR_CUSTOM_LIB_DIR=/usr/lib64/OpenEXR2 \
 	-DEXTRA_OSLEXEC_LIBRARIES="/usr/lib64/libpthread.so" \
 	-DEXTRA_OSLEXEC_LIBRARIES="/usr/lib64/libcurses.so" \
+        -DLLVM_CUSTOM=1 \
+	-DLLVM_STATIC=1 \
        -DLLVM_VERSION=${LLVM_VERSION} \
        -DLLVM_DIRECTORY=${LLVM_DIRECTORY}

@@ -6,8 +6,8 @@ if (NOT VERBOSE)
    set (Bison_FIND_QUIETLY true)
    set (Boost_FIND_QUIETLY true)
    set (Flex_FIND_QUIETLY true)
-    set (IlmBase_FIND_QUIETLY true)
    set (LLVM_FIND_QUIETLY true)
+    set (OpenEXR_FIND_QUIETLY true)
    set (OpenImageIO_FIND_QUIETLY true)
    set (Partio_FIND_QUIETLY true)
    set (PkgConfig_FIND_QUIETLY true)
@@ -52,16 +52,27 @@ endif ()
 ###########################################################################
 # IlmBase setup

-if (NOT ILMBASE_INCLUDE_DIR)
-    find_package (IlmBase REQUIRED)
+if (NOT OPENEXR_FOUND)
+    find_package (OpenEXR REQUIRED)
 endif ()

-include_directories ("${ILMBASE_INCLUDE_DIR}")
-include_directories ("${ILMBASE_INCLUDE_DIR}/OpenEXR")
+#OpenEXR 2.2 still has problems with importing ImathInt64.h unqualified
+#thus need for ilmbase/OpenEXR
+include_directories ("${OPENEXR_INCLUDE_DIR}"
+                     "${ILMBASE_INCLUDE_DIR}"
+                     "${ILMBASE_INCLUDE_DIR}/OpenEXR")

-macro (LINK_ILMBASE target)
-    target_link_libraries (${target} ${ILMBASE_LIBRARIES})
-endmacro ()
+if (${OPENEXR_VERSION} VERSION_LESS 2.0.0)
+    # OpenEXR 1.x had weird #include dirctives, this is also necessary:
+    include_directories ("${OPENEXR_INCLUDE_DIR}/OpenEXR")
+else ()
+    add_definitions (-DUSE_OPENEXR_VERSION2=1)
+endif ()
+
+if (NOT OpenEXR_FIND_QUIETLY)
+    message (STATUS "ILMBASE_INCLUDE_DIR = ${ILMBASE_INCLUDE_DIR}")
+    message (STATUS "ILMBASE_LIBRARIES = ${ILMBASE_LIBRARIES}")
+endif ()

 # end IlmBase setup
 ###########################################################################
@@ -169,9 +180,9 @@ endif()

 # try to find llvm-config, with a specific version if specified
 if(LLVM_DIRECTORY)
-  FIND_PROGRAM(LLVM_CONFIG llvm-config-${LLVM_VERSION} HINTS "${LLVM_DIRECTORY}/bin" NO_CMAKE_PATH)
+  FIND_PROGRAM(LLVM_CONFIG llvm-config-${LLVM_VERSION} HINTS "${LLVM_DIRECTORY}/bin" NO_DEFAULT_PATH)
  if(NOT LLVM_CONFIG)
-    FIND_PROGRAM(LLVM_CONFIG llvm-config HINTS "${LLVM_DIRECTORY}/bin" NO_CMAKE_PATH)
+    FIND_PROGRAM(LLVM_CONFIG llvm-config HINTS "${LLVM_DIRECTORY}/bin" NO_DEFAULT_PATH)
  endif()
 else()
  FIND_PROGRAM(LLVM_CONFIG llvm-config-${LLVM_VERSION})
@@ -1,260 +0,0 @@
-# Module to find IlmBase
-#
-# This module will first look into the directories defined by the variables:
-#   - ILMBASE_HOME, ILMBASE_VERSION, ILMBASE_LIB_AREA
-#
-# It also supports non-standard names for the library components.
-#
-# To use a custom IlmBase:
-#   - Set the variable ILMBASE_CUSTOM to True
-#   - Set the variable ILMBASE_CUSTOM_LIBRARIES to a list of the libraries to
-#     use, e.g. "SpiImath SpiHalf SpiIlmThread SpiIex"
-#   - Optionally set the variable ILMBASE_CUSTOM_INCLUDE_DIR to any
-#     particularly weird place that the OpenEXR/*.h files may be found
-#   - Optionally set the variable ILMBASE_CUSTOM_LIB_DIR to any
-#     particularly weird place that the libraries files may be found
-#
-# This module defines the following variables:
-#
-# ILMBASE_INCLUDE_DIR - where to find half.h, IlmBaseConfig.h, etc.
-# ILMBASE_LIBRARIES   - list of libraries to link against when using IlmBase.
-# ILMBASE_FOUND       - True if IlmBase was found.
-
-# Other standarnd issue macros
-include (FindPackageHandleStandardArgs)
-include (FindPackageMessage)
-include (SelectLibraryConfigurations)
-
-
-if( ILMBASE_USE_STATIC_LIBS )
-  set( _ilmbase_ORIG_CMAKE_FIND_LIBRARY_SUFFIXES ${CMAKE_FIND_LIBRARY_SUFFIXES})
-  if(WIN32)
-    set(CMAKE_FIND_LIBRARY_SUFFIXES .lib .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
-  else()
-    set(CMAKE_FIND_LIBRARY_SUFFIXES .a )
-  endif()
-endif()
-
-# Macro to assemble a helper state variable
-macro (SET_STATE_VAR varname)
-  set (tmp_ilmbaselibs ${ILMBASE_CUSTOM_LIBRARIES})
-  separate_arguments (tmp_ilmbaselibs)
-  set (tmp_lst
-    ${ILMBASE_CUSTOM} | ${tmp_ilmbaselibs} |
-    ${ILMBASE_HOME} | ${ILMBASE_VERSION} | ${ILMBASE_LIB_AREA}
-  )
-  set (${varname} "${tmp_lst}")
-  unset (tmp_ilmbaselibs)
-  unset (tmp_lst)
-endmacro ()
-
-# To enforce that find_* functions do not use inadvertently existing versions
-if (ILMBASE_CUSTOM)
-  set (ILMBASE_FIND_OPTIONS "NO_DEFAULT_PATH")
-endif ()
-
-# Macro to search for an include directory
-macro (PREFIX_FIND_INCLUDE_DIR prefix includefile libpath_var)
-  string (TOUPPER ${prefix}_INCLUDE_DIR tmp_varname)
-  find_path(${tmp_varname} ${includefile}
-    HINTS ${${libpath_var}}
-    PATH_SUFFIXES include
-    ${ILMBASE_FIND_OPTIONS}
-  )
-  if (${tmp_varname})
-    mark_as_advanced (${tmp_varname})
-  endif ()
-  unset (tmp_varname)
-endmacro ()
-
-
-# Macro to search for the given library and adds the cached
-# variable names to the specified list
-macro (PREFIX_FIND_LIB prefix libname libpath_var liblist_var cachelist_var)
-  string (TOUPPER ${prefix}_${libname} tmp_prefix)
-  # Handle new library names for OpenEXR 2.1 build via cmake
-  string(REPLACE "." "_" _ILMBASE_VERSION ${ILMBASE_VERSION})
-  string(SUBSTRING ${_ILMBASE_VERSION} 0 3 _ILMBASE_VERSION )
-  
-  find_library(${tmp_prefix}_LIBRARY_RELEASE
-    NAMES ${libname} ${libname}-${_ILMBASE_VERSION}
-    HINTS ${${libpath_var}}
-    PATH_SUFFIXES lib
-    ${ILMBASE_FIND_OPTIONS}
-  )
-  find_library(${tmp_prefix}_LIBRARY_DEBUG
-    NAMES ${libname}d ${libname}_d ${libname}debug ${libname}_debug
-    HINTS ${${libpath_var}}
-    PATH_SUFFIXES lib
-    ${ILMBASE_FIND_OPTIONS}
-  )
-  # Properly define ${tmp_prefix}_LIBRARY (cached) and ${tmp_prefix}_LIBRARIES
-  select_library_configurations (${tmp_prefix})
-  list (APPEND ${liblist_var} ${tmp_prefix}_LIBRARIES)
-
-  # Add to the list of variables which should be reset
-  list (APPEND ${cachelist_var}
-    ${tmp_prefix}_LIBRARY
-    ${tmp_prefix}_LIBRARY_RELEASE
-    ${tmp_prefix}_LIBRARY_DEBUG)
-  mark_as_advanced (
-    ${tmp_prefix}_LIBRARY
-    ${tmp_prefix}_LIBRARY_RELEASE
-    ${tmp_prefix}_LIBRARY_DEBUG)
-  unset (tmp_prefix)
-endmacro ()
-
-
-# Encode the current state of the external variables into a string
-SET_STATE_VAR (ILMBASE_CURRENT_STATE)
-
-# If the state has changed, clear the cached variables
-if (ILMBASE_CACHED_STATE AND
-    NOT ILMBASE_CACHED_STATE STREQUAL ILMBASE_CURRENT_STATE)
-
-  foreach (libvar ${ILMBASE_CACHED_VARS})
-    unset (${libvar} CACHE)
-  endforeach ()
-endif ()
-
-
-# Generic search paths
-set (IlmBase_generic_include_paths
-  ${ILMBASE_CUSTOM_INCLUDE_DIR}
-  /usr/include
-  /usr/include/${CMAKE_LIBRARY_ARCHITECTURE}
-  /usr/local/include
-  /sw/include
-  /opt/local/include)
-set (IlmBase_generic_library_paths
-  ${ILMBASE_CUSTOM_LIB_DIR}
-  /usr/lib
-  /usr/lib/${CMAKE_LIBRARY_ARCHITECTURE}
-  /usr/local/lib
-  /usr/local/lib/${CMAKE_LIBRARY_ARCHITECTURE}
-  /sw/lib
-  /opt/local/lib)
-
-# Search paths for the IlmBase files
-if (ILMBASE_HOME)
-  if (ILMBASE_VERSION)
-    set (IlmBase_include_paths
-      ${ILMBASE_HOME}/ilmbase-${ILMBASE_VERSION}/include
-      ${ILMBASE_HOME}/include/ilmbase-${ILMBASE_VERSION})
-    set (IlmBase_library_paths
-      ${ILMBASE_HOME}/ilmbase-${ILMBASE_VERSION}/lib
-      ${ILMBASE_HOME}/lib/ilmbase-${ILMBASE_VERSION})
-  endif()
-  list (APPEND IlmBase_include_paths ${ILMBASE_HOME}/include)
-  set (IlmBase_library_paths
-    ${ILMBASE_HOME}/lib
-    ${ILMBASE_HOME}/lib64
-    ${ILMBASE_LIB_AREA}
-    ${IlmBase_library_paths})
-endif ()
-list (APPEND IlmBase_include_paths ${IlmBase_generic_include_paths})
-list (APPEND IlmBase_library_paths ${IlmBase_generic_library_paths})
-
-# Locate the header files
-PREFIX_FIND_INCLUDE_DIR (IlmBase
-  OpenEXR/IlmBaseConfig.h IlmBase_include_paths)
-
-if (ILMBASE_INCLUDE_DIR)
-  # Get the version from config file, if not already set.
-  if (NOT ILMBASE_VERSION)
-    FILE(STRINGS "${ILMBASE_INCLUDE_DIR}/OpenEXR/IlmBaseConfig.h" ILMBASE_BUILD_SPECIFICATION
-         REGEX "^[ \t]*#define[ \t]+ILMBASE_VERSION_STRING[ \t]+\"[.0-9]+\".*$")
-
-    if(ILMBASE_BUILD_SPECIFICATION)
-      if (NOT IlmBase_FIND_QUIETLY)
-        message(STATUS "${ILMBASE_BUILD_SPECIFICATION}")
-      endif ()
-      string(REGEX REPLACE ".*#define[ \t]+ILMBASE_VERSION_STRING[ \t]+\"([.0-9]+)\".*"
-             "\\1" XYZ ${ILMBASE_BUILD_SPECIFICATION})
-      set("ILMBASE_VERSION" ${XYZ} CACHE STRING "Version of ILMBase lib")
-    else()
-      # Old versions (before 2.0?) do not have any version string, just assuming 2.0 should be fine though. 
-      message(WARNING "Could not determine ILMBase library version, assuming 2.0.")
-      set("ILMBASE_VERSION" "2.0" CACHE STRING "Version of ILMBase lib")
-    endif()
-  endif()
-endif ()
-
-
-if (ILMBASE_CUSTOM)
-  if (NOT ILMBASE_CUSTOM_LIBRARIES)
-    message (FATAL_ERROR "Custom IlmBase libraries requested but ILMBASE_CUSTOM_LIBRARIES is not set.")
-  endif()
-  set (IlmBase_Libraries ${ILMBASE_CUSTOM_LIBRARIES})
-  separate_arguments(IlmBase_Libraries)
-else ()
-#elseif (${ILMBASE_VERSION} VERSION_LESS "2.1")
-  set (IlmBase_Libraries Half Iex Imath IlmThread)
-#else ()
-#  string(REGEX REPLACE "([0-9]+)[.]([0-9]+).*" "\\1_\\2" _ilmbase_libs_ver ${ILMBASE_VERSION})
-#  set (IlmBase_Libraries Half Iex-${_ilmbase_libs_ver} Imath-${_ilmbase_libs_ver} IlmThread-${_ilmbase_libs_ver})
-endif ()
-
-
-# Locate the IlmBase libraries
-set (IlmBase_libvars "")
-set (IlmBase_cachevars "")
-foreach (ilmbase_lib ${IlmBase_Libraries})
-  PREFIX_FIND_LIB (IlmBase ${ilmbase_lib}
-    IlmBase_library_paths IlmBase_libvars IlmBase_cachevars)
-endforeach ()
-# Create the list of variables that might need to be cleared
-set (ILMBASE_CACHED_VARS
-  ILMBASE_INCLUDE_DIR ${IlmBase_cachevars}
-  CACHE INTERNAL "Variables set by FindIlmBase.cmake" FORCE)
-
-# Store the current state so that variables might be cleared if required
-set (ILMBASE_CACHED_STATE ${ILMBASE_CURRENT_STATE}
-  CACHE INTERNAL "State last seen by FindIlmBase.cmake" FORCE)
-
-# Link with pthreads if required
-if (NOT WIN32 AND EXISTS ${ILMBASE_INCLUDE_DIR}/OpenEXR/IlmBaseConfig.h)
-  file (STRINGS ${ILMBASE_INCLUDE_DIR}/OpenEXR/IlmBaseConfig.h
-    ILMBASE_HAVE_PTHREAD
-    REGEX "^[ \\t]*#define[ \\t]+HAVE_PTHREAD[ \\t]1[ \\t]*\$"
-  )
-  if (ILMBASE_HAVE_PTHREAD)
-    find_package (Threads)
-    if (CMAKE_USE_PTHREADS_INIT)
-      set (ILMBASE_PTHREADS ${CMAKE_THREAD_LIBS_INIT})
-    endif ()
-  endif ()
-endif ()
-
-# Use the standard function to handle ILMBASE_FOUND
-FIND_PACKAGE_HANDLE_STANDARD_ARGS (IlmBase DEFAULT_MSG
-  ILMBASE_INCLUDE_DIR ${IlmBase_libvars})
-
-if (ILMBASE_FOUND)
-  set (ILMBASE_LIBRARIES "")
-  foreach (tmplib ${IlmBase_libvars})
-    list (APPEND ILMBASE_LIBRARIES ${${tmplib}})
-  endforeach ()
-  list (APPEND ILMBASE_LIBRARIES ${ILMBASE_PTHREADS})
-  if (NOT IlmBase_FIND_QUIETLY)
-    FIND_PACKAGE_MESSAGE (ILMBASE
-      "Found IlmBase: ${ILMBASE_LIBRARIES}"
-      "[${ILMBASE_INCLUDE_DIR}][${ILMBASE_LIBRARIES}][${ILMBASE_CURRENT_STATE}]"
-      )
-  endif ()
-endif ()
-
-# Restore the original find library ordering
-if( ILMBASE_USE_STATIC_LIBS )
-  set(CMAKE_FIND_LIBRARY_SUFFIXES ${_ilmbase_ORIG_CMAKE_FIND_LIBRARY_SUFFIXES})
-endif()
-
-# Unset the helper variables to avoid pollution
-unset (ILMBASE_CURRENT_STATE)
-unset (IlmBase_include_paths)
-unset (IlmBase_library_paths)
-unset (IlmBase_generic_include_paths)
-unset (IlmBase_generic_library_paths)
-unset (IlmBase_libvars)
-unset (IlmBase_cachevars)
-unset (ILMBASE_PTHREADS)
@@ -0,0 +1,137 @@
+# Module to find OpenEXR.
+#
+# This module will set
+#   OPENEXR_FOUND          true, if found
+#   OPENEXR_INCLUDE_DIR    directory where headers are found
+#   OPENEXR_LIBRARIES      libraries for OpenEXR + IlmBase
+#   ILMBASE_LIBRARIES      libraries just IlmBase
+#   OPENEXR_VERSION        OpenEXR version (accurate for >= 2.0.0,
+#                              otherwise will just guess 1.6.1)
+#
+# Special inputs:
+#   OPENEXR_CUSTOM_INCLUDE_DIR - custom location of headers
+#   OPENEXR_CUSTOM_LIB_DIR - custom location of libraries
+#   OPENEXR_CUSTOM_LIB_PREFIX - special snowflake library prefix
+#   OPENEXR_CUSTOM_LIB_SUFFIX - special snowflake library suffix
+#
+
+# Other standarnd issue macros
+include (FindPackageHandleStandardArgs)
+include (SelectLibraryConfigurations)
+
+find_package (ZLIB REQUIRED)
+
+# List of likely places to find the headers -- note priority override of
+# OPENEXR_CUSTOM_INCLUDE_DIR and ${OPENEXR_HOME}/include.
+# ILMBASE is needed in case ilmbase an openexr are installed in separate
+# directories, like NixOS does
+set (GENERIC_INCLUDE_PATHS
+    ${OPENEXR_CUSTOM_INCLUDE_DIR}
+    ${OPENEXR_HOME}/include
+    ${ILMBASE_HOME}/include
+    /usr/include
+    /usr/include/${CMAKE_LIBRARY_ARCHITECTURE}
+    /usr/local/include
+    /sw/include
+    /opt/local/include )
+
+# List of likely places to find the libraries -- note priority override of
+# OPENEXR_CUSTOM_LIB_DIR and ${OPENEXR_HOME}/lib.
+set (GENERIC_LIBRARY_PATHS 
+    ${OPENEXR_CUSTOM_LIB_DIR}
+    ${OPENEXR_HOME}/lib
+    ${ILMBASE_HOME}/lib
+    /usr/lib
+    /usr/lib/${CMAKE_LIBRARY_ARCHITECTURE}
+    /usr/local/lib
+    /usr/local/lib/${CMAKE_LIBRARY_ARCHITECTURE}
+    /sw/lib
+    /opt/local/lib
+    $ENV{PROGRAM_FILES}/OpenEXR/lib/static )
+
+# Find the include file locations. We call find_path twice -- first using
+# only the custom paths, then if that fails, try the default paths only.
+# This seems to be the most robust way I can find to not get confused when
+# both system and custom libraries are present.
+find_path (ILMBASE_INCLUDE_PATH OpenEXR/IlmBaseConfig.h
+           PATHS ${GENERIC_INCLUDE_PATHS} NO_DEFAULT_PATH)
+find_path (ILMBASE_INCLUDE_PATH OpenEXR/IlmBaseConfig.h)
+find_path (OPENEXR_INCLUDE_PATH OpenEXR/OpenEXRConfig.h
+           PATHS ${GENERIC_INCLUDE_PATHS} NO_DEFAULT_PATH)
+find_path (OPENEXR_INCLUDE_PATH OpenEXR/OpenEXRConfig.h)
+
+# Handle request for static libs by altering CMAKE_FIND_LIBRARY_SUFFIXES.
+# We will restore it at the end of this file.
+set (_openexr_orig_suffixes ${CMAKE_FIND_LIBRARY_SUFFIXES})
+if (OpenEXR_USE_STATIC_LIBS)
+    if (WIN32)
+        set (CMAKE_FIND_LIBRARY_SUFFIXES .lib .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
+    else ()
+        set (CMAKE_FIND_LIBRARY_SUFFIXES .a)
+    endif ()
+endif ()
+
+# Look for the libraries themselves, for all the components. Like with the
+# headers, we do two finds -- first for custom locations, then for default.
+set (_openexr_components IlmThread IlmImf Imath Iex Half)
+foreach (COMPONENT ${_openexr_components})
+    string (TOUPPER ${COMPONENT} UPPERCOMPONENT)
+    set (FULL_COMPONENT_NAME ${OPENEXR_CUSTOM_LIB_PREFIX}${COMPONENT}${OPENEXR_CUSTOM_LIB_SUFFIX})
+    find_library (OPENEXR_${UPPERCOMPONENT}_LIBRARY ${FULL_COMPONENT_NAME}
+                  PATHS ${GENERIC_LIBRARY_PATHS} NO_DEFAULT_PATH)
+    find_library (OPENEXR_${UPPERCOMPONENT}_LIBRARY ${FULL_COMPONENT_NAME})
+endforeach ()
+
+#Half usually has no suffix
+find_library (OPENEXR_HALF_LIBRARY ${OPENEXR_CUSTOM_LIB_PREFIX}Half
+              PATHS ${GENERIC_LIBRARY_PATHS} NO_DEFAULT_PATH)
+find_library (OPENEXR_HALF_LIBRARY ${OPENEXR_CUSTOM_LIB_PREFIX}Half)
+
+# Link with pthreads if required
+find_package (Threads)
+if (CMAKE_USE_PTHREADS_INIT)
+    set (ILMBASE_PTHREADS ${CMAKE_THREAD_LIBS_INIT})
+endif ()
+
+# Set the FOUND, INCLUDE_DIR, and LIBRARIES variables.
+if (ILMBASE_INCLUDE_PATH AND OPENEXR_INCLUDE_PATH AND
+      OPENEXR_IMATH_LIBRARY AND OPENEXR_ILMIMF_LIBRARY AND
+      OPENEXR_IEX_LIBRARY AND OPENEXR_HALF_LIBRARY)
+    set (OPENEXR_FOUND TRUE)
+    set (ILMBASE_FOUND TRUE)
+    set (ILMBASE_INCLUDE_DIR ${ILMBASE_INCLUDE_PATH} CACHE STRING "The include paths needed to use IlmBase")
+    set (OPENEXR_INCLUDE_DIR ${OPENEXR_INCLUDE_PATH} CACHE STRING "The include paths needed to use OpenEXR")
+    set (ILMBASE_LIBRARIES ${OPENEXR_IMATH_LIBRARY} ${OPENEXR_IEX_LIBRARY} ${OPENEXR_HALF_LIBRARY} ${OPENEXR_ILMTHREAD_LIBRARY} ${ILMBASE_PTHREADS} CACHE STRING "The libraries needed to use IlmBase")
+    set (OPENEXR_LIBRARIES ${OPENEXR_ILMIMF_LIBRARY} ${ILMBASE_LIBRARIES} ${ZLIB_LIBRARY} CACHE STRING "The libraries needed to use OpenEXR")
+endif ()
+
+# Try to figure out version number
+if (EXISTS "${OPENEXR_INCLUDE_DIR}/OpenEXR/ImfMultiPartInputFile.h")
+    # Must be at least 2.0
+    file(STRINGS "${OPENEXR_INCLUDE_DIR}/OpenEXR/OpenEXRConfig.h" TMP REGEX "^#define OPENEXR_VERSION_STRING .*$")
+    string (REGEX MATCHALL "[0-9]+[.0-9]+" OPENEXR_VERSION ${TMP})
+else ()
+    # Assume an old one, predates 2.x that had versions
+    set (OPENEXR_VERSION 1.6.1)
+endif ()
+
+find_package_handle_standard_args (OpenEXR
+    REQUIRED_VARS ILMBASE_INCLUDE_PATH OPENEXR_INCLUDE_PATH
+                  OPENEXR_IMATH_LIBRARY OPENEXR_ILMIMF_LIBRARY
+                  OPENEXR_IEX_LIBRARY OPENEXR_HALF_LIBRARY
+    VERSION_VAR   OPENEXR_VERSION
+    )
+
+MARK_AS_ADVANCED(
+    ILMBASE_INCLUDE_DIR
+    OPENEXR_INCLUDE_DIR
+    ILMBASE_LIBRARIES
+    OPENEXR_LIBRARIES
+    OPENEXR_ILMIMF_LIBRARY
+    OPENEXR_IMATH_LIBRARY
+    OPENEXR_IEX_LIBRARY
+    OPENEXR_HALF_LIBRARY
+    OPENEXR_VERSION)
+
+# Restore the original CMAKE_FIND_LIBRARY_SUFFIXES
+set (CMAKE_FIND_LIBRARY_SUFFIXES ${_openexr_orig_suffixes})
@@ -62,7 +62,7 @@ Editor: Larry Gritz \\
 \emph{lg@imageworks.com}
 }
 \date{{\large Date: 14 Jan, 2016 \\
-%~ (with corrections, 14 Jan 2016)
+ (with corrections, 5 Jul 2016)
 }
 \bigskip
 \bigskip
@@ -472,15 +472,15 @@ overridden with an instance-specific value of {\cf 2.2}.



-\subsection*{Shader groups and layers}
+\subsection*{Shader groups, layers, and connections}

-A \emph{shader group} is an ordered sequence of individual shaders
-called \emph{layers} that are executed in turn.  Output parameters of an
-earlier-executed layer may be \emph{connected} to an input parameter of
-a later-executed layer.  This connected network of layers is sometimes
-called a \emph{shader network} or a \emph{shader DAG} (directed acyclic
-graph).  Of course, it is fine for the shader group to consist of a
-single shader layer.
+A \emph{shader group}\index{shader group} is an ordered sequence of
+individual shaders called \emph{layers}\index{layer}\index{shader layer}
+that are executed in turn.  Output parameters of an earlier-executed layer
+may be \emph{connected} to an input parameter of a later-executed layer.
+This connected network of layers is sometimes called a \emph{shader network}
+or a \emph{shader DAG} (directed acyclic graph).  Of course, it is fine for
+the shader group to consist of a single shader layer.

 Below is a schematic showing how several shader instances may be
 connected to form a shader group.
@@ -517,7 +517,7 @@ it might be done.}:
 \end{code}

 \noindent Or, expressed as serialized text (as detailed in
-Appendix~\ref{chap:serialize}):
+Chapter~\ref{sec:serialize}):

 \begin{code}
    param string name "rings.tx" ;
@@ -535,7 +535,18 @@ Appendix~\ref{chap:serialize}):
    connect gam2.Cout wood1.grain ;
 \end{code}

-
+The rules for which data types may be connected are generally the same as
+the rules determining which variables may be assigned to each other in OSL
+source code:
+\begin{itemize}
+\item {\cf source} and {\cf dest} are the same data type.
+\item {\cf source} and {\cf dest} are both \emph{triples} (\color, \point,
+\vector, or \normal), even if they are not the same kind of triple.
+\item {\cf source} is an {\cf int} and {\cf dest} is a {\cf float}.
+\item {\cf source} is a {\cf float} or {\cf int} and {\cf dest} is a
+\emph{triple} (the scalar value will be replicated for all three components
+of the triple).
+\end{itemize}

 \subsection*{Geometric primitives}

@@ -547,7 +558,8 @@ polygons, and curves.  The exact set of supported primitives may vary
 from renderer to renderer.

 Each geometric primitive carries around a set of named \emph{primitive
-  variables}.  Nearly all shape types will have, among their primitive
+variables} (also sometimes called \emph{interpolated values} or
+\emph{user data}).  Nearly all shape types will have, among their primitive
 variables, control point positions that, when interpolated, actually
 designate the shape.  Some shapes will also allow the specification of
 normals or other shape-specific data.  Arbitrary user data may also be
@@ -842,6 +854,7 @@ uniform union unsigned varying virtual volatile
 \indexapi{\#else}
 \indexapi{\#endif}
 \indexapi{\#include}
+\indexapi{\#pragma}

 Shader source code is passed through a standard C preprocessor as a
 first step in parsing.  
@@ -864,6 +877,7 @@ directives, including:
    #else
    #endif
    #include
+    #pragma once
 \end{code}


@@ -1450,7 +1464,8 @@ the {\cf int} type is renderer-dependent, but is guaranteed to be at
 least 32 bits.

 Integer constants are constructed the same way as in C.  The following
-are examples of {\cf int} constants: {\cf 1}, {\cf -32}, etc.
+are examples of {\cf int} constants: {\cf 1}, {\cf -32}, etc. Integer
+constants may be specified as hexidecimal, for example: {\cf 0x01cf}.

 Unlike C, no unsigned, bool, char, short, or long types are supplied.
 This is to simplify the process of writing shaders (as well as
@@ -2573,7 +2588,7 @@ loop:

 \indexapi{for}
 As in C++, a {\cf for} loop's initialization may contain variable
-declarations and initializations, which are scoped locally to {\cf for}
+declarations and initializations, which are scoped locally to the {\cf for}
 loop itself.  For example,

 \begin{code}
@@ -2656,9 +2671,9 @@ P/\partial u$ and $\partial P/\partial v$ tangent to the surface at \P.  \\
 %         {\cf Pl}). \\
 %\hline
 %\color\ {\ce Cl} & Incoming light color at P. \\
-%\hline
+\hline
 \point\ {\ce Ps} & Position at which the light is being queried 
-    (currently only used for light attentuation shaders) \\
+    (currently only used for light attenuation shaders) \\
 %\\
 %\hline
 %\normal\ {\ce Ns} & Axis about which the light is being queried
@@ -2729,6 +2744,7 @@ P/\partial u$ and $\partial P/\partial v$ tangent to the surface at \P.  \\
 {\cf Ng}      & R     & R  &    \\
 {\cf dPdu}    & R     & R  &    \\
 {\cf dPdv}    & R     & R  &    \\
+{\cf Ps}      &       &    & R   \\
 {\cf u, v}    & R     & R  & R   \\
 {\cf time}    & R     & R  & R       \\
 {\cf dtime}   & R     & R  & R       \\
@@ -3417,7 +3433,7 @@ derivatives.  Gabor noise allows several optional parameters to the
 "direction", <vector>}
 \vspace{12pt}
 If {\cf anisotropic} is 0 (the default), Gabor noise
-will be isotropic.  If {\cf anisotripic} is 1, the Gabor noise will
+will be isotropic.  If {\cf anisotropic} is 1, the Gabor noise will
 be anisotropic with the 3D frequency given by the {\cf direction} 
 vector (which defaults to {\cf (1,0,0)}).  If {\cf anisotropic} is
 2, a hybrid mode will be used which is anisotropic along the
@@ -4273,7 +4289,7 @@ given {\cf data} arrays.  If the optional {\cf sort} parameter is present
 and is nonzero, the ordering of the points
 found will be sorted by distance from {\cf pos}, from closest to farthest;
 otherwise, the results are guaranteed to be the {\cf maxpoints} closest
-to {\cf pos}, but not necessarily softed by distance (this may be
+to {\cf pos}, but not necessarily sorted by distance (this may be
 faster for some implementations than when sorted results are
 required).  The return value is the number of points returned, ranging
 from 0 (nothing found in the neighborhood) to the lesser of {\cf maxpoints}
@@ -4530,7 +4546,7 @@ the previous function.
 \apiend


-\apiitem{\colorclosure\ \colorclosure\ {\ce reflection} (normal N, float eta)}
+\apiitem{\colorclosure\ {\ce reflection} (normal N, float eta)}
 \indexapi{reflection()}

 Returns a \colorclosure that represents sharp mirror-like reflection
@@ -4714,7 +4730,8 @@ int {\ce getattribute} (string object, string name, output \emph{type} destinati
 int {\ce getattribute} (string object, string name, int arrayindex, \\
 \bigspc\spc\spc output \emph{type} destination)}
 \indexapi{getattribute()}
-Searches for the renderer attribute with the given \emph{name}.
+Retrieves a named renderer attribute or the value of an interpolated
+geometric variable.
 If an object is explicitly named, that is the only place that will be
 searched (\qkw{global} means the global scene-wide attributes).  For the
 forms of the function with no object name, or if the object name is the
@@ -5077,7 +5094,12 @@ nothing (empty, no token).

 <digit-sequence> ::= <digit> \{ <digit> \}

+<hexdigit> ::= <digit> | "a" | "A" | "b" | "B" | "c" | "C" | "d" | "D" | "e" | "E" | "f" | "F"
+
+<hexdigit-sequence> ::= <hexdigit> \{ <hexdigit> \}
+
 <integer> ::= <sign> <digit-sequence>
+\alt <sign> "0x" <hexdigit-sequence>

 <floating-point> ::= <digit-sequence> <decimal-part-opt> <exponent-opt>
 \alt <decimal-part> <exponent-opt>
@@ -123,6 +123,9 @@ public:
    ///    int lazyunconnected    Run layers lazily even if they have no
    ///                              output connections (1). For debugging.
    ///    int lazy_userdata      Retrieve userdata lazily (0).
+    ///    int userdata_isconnected  Should lockgeom=0 params (that may
+    ///                              receive userdata) return true from
+    ///                              isconnected()? (0)
    ///    int greedyjit          Optimize and compile all shaders up front,
    ///                              versus only as needed (0).
    ///    int lockgeom           Default 'lockgeom' value for shader params
@@ -20,8 +20,8 @@ else ()
    ADD_LIBRARY ( oslcomp SHARED ${liboslcomp_srcs} )
 endif ()

-TARGET_LINK_LIBRARIES ( oslcomp ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
-LINK_ILMBASE ( oslcomp )
+TARGET_LINK_LIBRARIES ( oslcomp ${OPENIMAGEIO_LIBRARY} ${ILMBASE_LIBRARIES}
+                       ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})

 INSTALL ( TARGETS oslcomp RUNTIME DESTINATION bin LIBRARY DESTINATION lib ARCHIVE DESTINATION lib )

@@ -260,10 +260,18 @@ ASTfunction_declaration::add_meta (ASTNode *meta)
        Symbol *metasym = metavar->sym();
        if (metasym->name() == "builtin") {
            m_is_builtin = true;
-            if (func()->typespec().is_closure()) // It is a builtin closure
+            if (func()->typespec().is_closure())  { // It is a builtin closure
                // Force keyword arguments at the end
                func()->argcodes(ustring(std::string(func()->argcodes().c_str()) + "."));
-
+            }
+            // For built-in functions, if any of the params are output,
+            // also automatically mark it as readwrite_special_case.
+            for (ASTNode *f = formals().get(); f; f = f->nextptr()) {
+                ASSERT (f->nodetype() == variable_declaration_node);
+                ASTvariable_declaration *v = (ASTvariable_declaration *)f;
+                if (v->is_output())
+                    func()->readwrite_special_case (true);
+            }
        }
        else if (metasym->name() == "derivs")
            func()->takes_derivs (true);
@@ -446,7 +454,8 @@ ASTindex::ASTindex (OSLCompilerImpl *comp, ASTNode *expr, ASTNode *index)
            expr->nodetype() == structselect_node);
    if (expr->typespec().is_array())       // array dereference
        m_typespec = expr->typespec().elementtype();
-    else if (expr->typespec().is_triple()) // component access
+    else if (!expr->typespec().is_closure() &&
+             expr->typespec().is_triple()) // component access
        m_typespec = TypeDesc::FLOAT;
    else {
        error ("indexing into non-array or non-component type");
@@ -900,6 +909,7 @@ ASTfunction_call::ASTfunction_call (OSLCompilerImpl *comp, ustring name,
    }
    if (m_sym->symtype() != SymTypeFunction) {
        error ("'%s' is not a function", name.c_str());
+        m_sym = NULL;
        return;
    }
 }
@@ -663,12 +663,15 @@ ASTvariable_declaration::param_one_default_literal (const Symbol *sym,
                val = val->next();
                completed = false;
            }
+            bool one_arg = (val && ! val->nextptr());
            for (int c = 0;  c < 3;  ++c) {
                if (val.get())
                    ++nargs;
                if (val.get() && val->nodetype() == ASTNode::literal_node) {
                    f[c] = ((ASTliteral *)val.get())->floatval ();
                    val = val->next();
+                } else if (c > 0 && one_arg) {
+                    f[c] = f[0];
                } else {
                    f[c] = 0;
                    completed = false;
@@ -699,6 +702,11 @@ ASTvariable_declaration::param_one_default_literal (const Symbol *sym,
            ASTNode::ref val = ctr->args();
            float f[16];
            int nargs = 0;
+            DASSERT (val.get());
+            if (val->nodetype() == ASTNode::literal_node && val->typespec().is_string()) {
+                val = val->next();
+                completed = false;
+            }
            for (int c = 0;  c < 16;  ++c) {
                if (val.get())
                    ++nargs;
@@ -41,6 +41,11 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 %option noyywrap
 %option nounput noinput

+/* never-interactive fixes a Windows compatibility problem where the
+ * lexer emits isatty calls that don't exist.
+ */
+%option never-interactive
+
 /* Option 'prefix' creates a C++ lexer with the given prefix, so that
  * we can link with other flex-generated lexers in the same application
  * without name conflicts.
@@ -59,6 +64,7 @@ ALPHA           [A-Za-z]
 DIGIT           [0-9]
 /* Integer literal */
 INTEGER         {DIGIT}+
+HEXINTEGER      0[xX][0-9a-fA-F]+
 /* floating point literal (E, FLT1, FLT2, FLT3 are just helpers)
  * NB: we don't allow leading +/- due to ambiguity between
  * whether "a-0.5" is really "a -0.5" or "a - 0.5".  Resolve this
@@ -91,6 +97,7 @@ CPLUSCOMMENT    \/\/.*\n
 #include <vector>
 #include <string>

+#include <OpenImageIO/thread.h>
 #include "oslcomp_pvt.h"

 using namespace OSL;
@@ -106,6 +113,8 @@ using namespace OSL::pvt;

 #ifdef _MSC_VER
 #define strtoll _strtoi64
+#include <io.h>
+#define isatty _isatty
 #endif
 #endif

@@ -196,6 +205,22 @@ void preprocess (const char *yytext);
                            SETLINE;
                            return INT_LITERAL;
                        }
+{HEXINTEGER}            {
+                            long long llval = strtoll (yytext, (char**)NULL, 16);
+                            // we do not detect overflow when the value is INT_MAX+1,
+                            // because negation happens later and -(INT_MAX+1) == INT_MIN
+                            if (llval > ((long long)UINT_MAX)+1) {
+                                oslcompiler->error (oslcompiler->filename(),
+                                                    oslcompiler->lineno(),
+                                                    "integer overflow, value must be between %d and %d.",
+                                                    INT_MIN, INT_MAX);
+                            }
+                            yylval.i = (int)llval;
+                            SETLINE;
+                            return INT_LITERAL;
+                        }
+
+

 {FLT}                   {
                            yylval.f = atof (yytext);
@@ -122,6 +122,19 @@ ASTvariable_declaration::typecheck (TypeSpec expected)

    typecheck_initlist (init, m_typespec, m_name.c_str());

+    // Warning to catch confusing comma operator in variable initializers.
+    // One place this comes up is when somebody forgets the proper syntax
+    // for constructors, for example
+    //     color x = (a, b, c);   // Sets x to (c,c,c)!
+    // when they really meant
+    //     color x = color(a, b, c);
+    if (init->nodetype() == comma_operator_node && !typespec().is_closure() &&
+        (typespec().is_triple() || typespec().is_matrix())) {
+        warning ("Comma operator is very confusing here. "
+                 "Did you mean to use a constructor: %s = %s(...)?",
+                 m_name.c_str(), typespec().c_str());
+    }
+
    return m_typespec;
 }

@@ -432,6 +445,19 @@ ASTassign_expression::typecheck (TypeSpec expected)
        return TypeSpec();
    }

+    // Warning to catch confusing comma operator in assignment.
+    // One place this comes up is when somebody forgets the proper syntax
+    // for constructors, for example
+    //     color x = (a, b, c);   // Sets x to (c,c,c)!
+    // when they really meant
+    //     color x = color(a, b, c);
+    if (expr()->nodetype() == comma_operator_node && !vt.is_closure() &&
+        (vt.is_triple() || vt.is_matrix())) {
+        warning ("Comma operator is very confusing here. "
+                 "Did you mean to use a constructor: = %s(...)?",
+                 vt.c_str());
+    }
+
    return m_typespec = vt;
 }

@@ -993,7 +1019,8 @@ ASTfunction_call::typecheck_builtin_specialcase ()
            argwriteonly (1);
            argwriteonly (2);
        } else if (m_name == "getattribute" || m_name == "getmessage" ||
-                   m_name == "gettextureinfo" || m_name == "dict_value") {
+                   m_name == "gettextureinfo" || m_name == "getmatrix" ||
+                   m_name == "dict_value") {
            // these all write to their last argument
            argwriteonly ((int)listlength(args()));
        } else if (m_name == "pointcloud_get") {
@@ -1167,8 +1194,12 @@ ASTfunction_call::typecheck (TypeSpec expected)
        actualargs += arg->typespec().string();
    }

+    if (choices.size())
        error ("No matching function call to '%s (%s)'\n    Candidates are:\n%s",
               m_name.c_str(), actualargs.c_str(), choices.c_str());
+    else
+        error ("No matching function call to '%s (%s)'",
+               m_name.c_str(), actualargs.c_str());
    return TypeSpec();
 }

@@ -69,9 +69,7 @@ MACRO ( LLVM_COMPILE llvm_src srclist )
    foreach (def ${CURRENT_DEFINITIONS})
        set (LLVM_COMPILE_FLAGS ${LLVM_COMPILE_FLAGS} "-D${def}")
    endforeach()
-    if (USE_SIMD)
-    	set (LLVM_COMPILE_FLAGS ${LLVM_COMPILE_FLAGS} "-m${USE_SIMD}")
-    endif ()
+    set (LLVM_COMPILE_FLAGS ${LLVM_COMPILE_FLAGS} ${OSL_SIMD_FLAGS} ${OSL_CSTD_FLAGS})

    # Figure out what program we will use to make the bitcode.
    if (NOT LLVM_BC_GENERATOR)
@@ -116,6 +114,7 @@ MACRO ( LLVM_COMPILE llvm_src srclist )
          "-I${ILMBASE_INCLUDE_DIR}"
          "-I${Boost_INCLUDE_DIRS}"
          -DOSL_COMPILING_TO_BITCODE=1
+          -Wno-deprecated-register
          -O3 -fno-math-errno -S -emit-llvm -o ${llvm_asm} ${llvm_src}
      COMMAND "${LLVM_DIRECTORY}/bin/llvm-as" -f -o ${llvm_bc} ${llvm_asm}
      COMMAND "${CMAKE_CURRENT_SOURCE_DIR}/serialize-bc.bash" ${llvm_bc} ${llvm_bc_cpp}
@@ -142,14 +141,14 @@ endif ()

 TARGET_LINK_LIBRARIES ( oslexec
                        ${VISIBILITY_COMMAND} ${VISIBILITY_MAP_COMMAND}
-                        ${OPENIMAGEIO_LIBRARY} ${PUGIXML_LIBRARIES}
+                        ${OPENIMAGEIO_LIBRARY} ${ILMBASE_LIBRARIES}
+                        ${PUGIXML_LIBRARIES}
                        ${PARTIO_LIBRARIES} ${ZLIB_LIBRARIES}
                        ${Boost_LIBRARIES} ${CMAKE_DL_LIBS}
                        ${LLVM_LIBRARY} ${LLVM_MCJIT_LIBRARY}
                        ${LLVM_LIBRARIES} ${LLVM_LDFLAGS}
                        ${EXTRA_OSLEXEC_LIBRARIES})
 ADD_DEPENDENCIES (oslexec "${CMAKE_CURRENT_SOURCE_DIR}/liboslexec.map")
-LINK_ILMBASE ( oslexec )

 INSTALL ( TARGETS oslexec RUNTIME DESTINATION bin LIBRARY DESTINATION lib ARCHIVE DESTINATION lib )

@@ -157,11 +156,9 @@ INSTALL ( TARGETS oslexec RUNTIME DESTINATION bin LIBRARY DESTINATION lib ARCHIV
 if (OSL_BUILD_TESTS)
    add_executable (accum_test accum_test.cpp)
    target_link_libraries ( accum_test oslexec oslcomp ${Boost_LIBRARIES} ${CMAKE_DL_LIBS} ${EXTRA_OSLEXEC_LIBRARIES} )
-    link_ilmbase (accum_test)
    add_test (unit_accum "${CMAKE_BINARY_DIR}/src/liboslexec/accum_test")

    add_executable (llvmutil_test llvmutil_test.cpp)
    target_link_libraries ( llvmutil_test oslexec oslcomp ${Boost_LIBRARIES} ${CMAKE_DL_LIBS} ${EXTRA_OSLEXEC_LIBRARIES} )
-    link_ilmbase (llvmutil_test)
    add_test (unit_llvmutil "${CMAKE_BINARY_DIR}/src/liboslexec/llvmutil_test")
 endif ()
@@ -455,9 +455,6 @@ ShadingContext::free_dict_resources ()
 }


-#define USTR(cstr) (*((ustring *)&cstr))
-#define TYPEDESC(x) (*(TypeDesc *)&x)
-

 OSL_SHADEOP int osl_dict_find_iis (void *sg_, int nodeID, void *query)
 {
@@ -214,9 +214,13 @@ ShaderInstance::parameters (const ParamValueList &params)

    m_instoverrides.resize (std::max (0, lastparam()));

-    // Set the initial lockgeom on the instoverrides, based on the master.
-    for (int i = 0, e = (int)m_instoverrides.size(); i < e; ++i)
-        m_instoverrides[i].lockgeom (master()->symbol(i)->lockgeom());
+    // Set the initial lockgeom and dataoffset on the instoverrides, based
+    // on the master.
+    for (int i = 0, e = (int)m_instoverrides.size(); i < e; ++i) {
+        Symbol *sym = master()->symbol(i);
+        m_instoverrides[i].lockgeom (sym->lockgeom());
+        m_instoverrides[i].dataoffset (sym->dataoffset());
+    }

    BOOST_FOREACH (const ParamValue &p, params) {
        if (p.name().size() == 0)
@@ -257,6 +261,9 @@ ShaderInstance::parameters (const ParamValueList &params)
                             p.interp() == ParamValue::INTERP_CONSTANT);
            so->lockgeom (lockgeom);

+            DASSERT (so->dataoffset() == sm->dataoffset());
+            so->dataoffset (sm->dataoffset());
+
            if (paramtype.arraylen < 0) {
                // An array of definite size was supplied to a parameter
                // that was an array of indefinite size. Magic! The trick
@@ -460,6 +467,7 @@ ShaderInstance::copy_code_from_master (ShaderGroup &group)
                si->valuesource (m_instoverrides[i].valuesource());
                si->connected_down (m_instoverrides[i].connected_down());
                si->lockgeom (m_instoverrides[i].lockgeom());
+                si->dataoffset (m_instoverrides[i].dataoffset());
                si->data (param_storage(i));
            }
            if (shadingsys().is_renderer_output (layername(), si->name(), &group)) {
@@ -2023,8 +2023,8 @@ llvm_gen_texture_options (BackendLLVM &rop, int opnum,

        Symbol &Val (*rop.opargsym(op,a));
        TypeDesc valtype = Val.typespec().simpletype ();
-        const int *ival = Val.typespec().is_int() ? (const int *)Val.data() : NULL;
-        const float *fval = Val.typespec().is_float() ? (const float *)Val.data() : NULL;
+        const int *ival = Val.typespec().is_int() && Val.is_constant() ? (const int *)Val.data() : NULL;
+        const float *fval = Val.typespec().is_float() && Val.is_constant() ? (const float *)Val.data() : NULL;

 #define PARAM_INT(paramname)                                            \
        if (name == Strings::paramname && valtype == TypeDesc::INT)   { \
@@ -150,7 +150,7 @@ initialize_llvm_helper_function_map ()
    if (llvm_helper_function_map_initialized)
        return;  // already done
    spin_lock lock (llvm_helper_function_map_mutex);
-    if (llvm_helper_function_map_initialized())
+    if (llvm_helper_function_map_initialized)
        return;
 #define DECL(name,signature) \
    llvm_helper_function_map[#name] = HelperFuncRecord(signature,name); \
@@ -750,9 +750,10 @@ BackendLLVM::build_llvm_instance (bool groupentry)
    // Note that the GroupData* is passed as a void*.
    std::string unique_layer_name = Strutil::format ("%s_%d", inst()->layername(), inst()->id());

+    bool is_entry_layer = group().is_entry_layer(layer());
    ll.current_function (
           ll.make_function (unique_layer_name,
-                             !inst()->entry_layer(), // fastcall for non-entry layer functions
+                             !is_entry_layer, // fastcall for non-entry layer functions
                             ll.type_void(), // return type
                             llvm_type_sg_ptr(), llvm_type_groupdata_ptr()));

@@ -767,7 +768,7 @@ BackendLLVM::build_llvm_instance (bool groupentry)
    ll.new_builder (entry_bb);

    llvm::Value *layerfield = layer_run_ref(layer_remap(layer()));
-    if (inst()->entry_layer()) {
+    if (is_entry_layer && ! group().is_last_layer(layer())) {
        // For entry layers, we need an extra check to see if it already
        // ran. If it has, do an early return. Otherwise, set the 'ran' flag
        // and then run the layer.
@@ -791,9 +792,11 @@ BackendLLVM::build_llvm_instance (bool groupentry)
        llvm_gen_debug_printf (Strutil::format("enter layer %d %s %s",
                               this->layer(), inst()->layername(), inst()->shadername()));
    // Mark this layer as executed
+    if (! group().is_last_layer(layer())) {
        ll.op_store (ll.constant_bool(true), layerfield);
        if (shadingsys().countlayerexecs())
            ll.call_function ("osl_incr_layers_executed", sg_void_ptr());
+    }

    // Setup the symbols
    m_named_values.clear ();
@@ -91,7 +91,15 @@ examples), as you are just coding in C++, but there are some rules:
 */


+// Some gcc versions on some platforms seem to have max_align_t missing from
+// their <cstddef>. Putting this here appears to make it build cleanly on
+// those platforms while not hurting anything elsewhere.
+namespace {
+typedef long double max_align_t;
+}
+
 #include <iostream>
+#include <cstddef>

 #include "OSL/oslconfig.h"
 #include "OSL/shaderglobals.h"
@@ -138,7 +146,6 @@ void * __dso_handle = 0; // necessary to avoid linkage issues in bitcode
 #define DVEC(x) (*(Dual2<Vec3> *)x)
 #define COL(x) (*(Color3 *)x)
 #define DCOL(x) (*(Dual2<Color3> *)x)
-#define TYPEDESC(x) (*(TypeDesc *)&x)


 #ifndef OSL_SHADEOP
@@ -643,14 +650,14 @@ OSL_SHADEOP void osl_transformn_vmv(void *result, void* M_, void* v_)
 {
   const Vec3 &v = VEC(v_);
   const Matrix44 &M = MAT(M_);
-   M.inverse().transpose().multDirMatrix (v, VEC(result));
+   M.inverse().transposed().multDirMatrix (v, VEC(result));
 }

 OSL_SHADEOP void osl_transformn_dvmdv(void *result, void* M_, void* v_)
 {
   const Dual2<Vec3> &v = DVEC(v_);
   const Matrix44    &M = MAT(M_);
-   multDirMatrix (M.inverse().transpose(), v, DVEC(result));
+   multDirMatrix (M.inverse().transposed(), v, DVEC(result));
 }


@@ -823,5 +830,3 @@ OSL_SHADEOP int osl_raytype_bit (void *sg_, int bit)
    ShaderGlobals *sg = (ShaderGlobals *)sg_;
    return (sg->raytype & bit) != 0;
 }
-
-
@@ -435,6 +435,10 @@ OSOReaderToMaster::hint (string_view hintstring)
            m_master->m_ops.back().argread(i, *str == 'r' || *str =='W');
        }
        ASSERT(m_nargs == i);
+        // Fix old bug where oslc forgot to mark getmatrix last arg as write
+        static ustring getmatrix("getmatrix");
+        if (m_master->m_ops.back().opname() == getmatrix)
+            m_master->m_ops.back().argwrite(m_nargs-1, true);
    }
    if (extract_prefix(h, "%argderivs{")) {
        while (1) {
@@ -227,13 +227,13 @@ inline void osl_transformv_dvmdv(void *result, const Matrix44 &M, void* v_)
 inline void osl_transformn_vmv(void *result, const Matrix44 &M, void* v_)
 {
   const Vec3 &v = VEC(v_);
-   M.inverse().transpose().multDirMatrix (v, VEC(result));
+   M.inverse().transposed().multDirMatrix (v, VEC(result));
 }

 inline void osl_transformn_dvmdv(void *result, const Matrix44 &M, void* v_)
 {
   const Dual2<Vec3> &v = DVEC(v_);
-   multDirMatrix (M.inverse().transpose(), v, DVEC(result));
+   multDirMatrix (M.inverse().transposed(), v, DVEC(result));
 }


@@ -50,15 +50,13 @@ OSL_NAMESPACE_ENTER
 namespace pvt {


-#define USTR(cstr) (*((ustring *)&cstr))
-
 OSL_SHADEOP void
 osl_setmessage (ShaderGlobals *sg, const char *name_, long long type_, void *val, int layeridx, const char* sourcefile_, int sourceline)
 {
    const ustring &name (USTR(name_));
    const ustring &sourcefile (USTR(sourcefile_));
    // recreate TypeDesc -- we just crammed it into an int!
-    TypeDesc type (*(TypeDesc *)&type_);
+    TypeDesc type = TYPEDESC(type_);
    bool is_closure = (type.basetype == TypeDesc::UNKNOWN); // secret code for closure
    if (is_closure)
        type.basetype = TypeDesc::PTR;  // for closures, we store a pointer
@@ -105,7 +103,7 @@ osl_getmessage (ShaderGlobals *sg, const char *source_, const char *name_,
    const ustring &sourcefile (USTR(sourcefile_));

    // recreate TypeDesc -- we just crammed it into an int!
-    TypeDesc type (*(TypeDesc *)&type_);
+    TypeDesc type = TYPEDESC(type_);
    bool is_closure = (type.basetype == TypeDesc::UNKNOWN); // secret code for closure
    if (is_closure)
        type.basetype = TypeDesc::PTR;  // for closures, we store a pointer
@@ -154,8 +154,8 @@ OSL_SHADEOP void  osl_spline_dvdfv(void *out, const char *spline_, void *x,
                                   Vec3 *knots, int knot_count, int knot_arraylen)
 {
   const Spline::SplineBasis *spline = Spline::getSplineBasis(USTR(spline_));
-   Spline::spline_evaluate<Vec3, float, Vec3, Vec3, false>
-      (spline, *(Vec3 *)out, *(float *)x, knots, knot_count, knot_arraylen);
+   Spline::spline_evaluate<Dual2<Vec3>, Dual2<float>, Vec3, Vec3, false>
+      (spline, DVEC(out), DFLOAT(x), knots, knot_count, knot_arraylen);
 }

 OSL_SHADEOP void  osl_spline_dvfdv(void *out, const char *spline_, void *x, 
@@ -215,7 +215,7 @@ struct AttributeNeeded {
 #define DVEC(x) (*(Dual2<Vec3> *)x)
 #define COL(x) (*(Color3 *)x)
 #define DCOL(x) (*(Dual2<Color3> *)x)
-#define TYPEDESC(x) (*(TypeDesc *)&x)
+#define TYPEDESC(x) OIIO::bit_cast<long long,TypeDesc>(x)



@@ -562,6 +562,7 @@ public:
    int max_warnings_per_thread() const { return m_max_warnings_per_thread; }
    bool countlayerexecs() const { return m_countlayerexecs; }
    bool lazy_userdata () const { return m_lazy_userdata; }
+    bool userdata_isconnected () const { return m_userdata_isconnected; }

    ustring commonspace_synonym () const { return m_commonspace_synonym; }

@@ -705,6 +706,7 @@ private:
    bool m_lazyglobals;                   ///< Run lazily even if globals write?
    bool m_lazyunconnected;               ///< Run lazily even if not connected?
    bool m_lazy_userdata;                 ///< Retrieve userdata lazily?
+    bool m_userdata_isconnected;          ///< Userdata params isconnected()?
    bool m_clearmemory;                   ///< Zero mem before running shader?
    bool m_debugnan;                      ///< Root out NaN's?
    bool m_debug_uninit;                  ///< Find use of uninitialized vars?
@@ -1433,12 +1435,16 @@ public:

    int num_entry_layers () const { return m_num_entry_layers; }

+    bool is_last_layer (int layer) const {
+        return layer == nlayers()-1;
+    }
+
    /// Is the given layer an entry point? It is if explicitly tagged as
    /// such, or if no layers are so tagged then the last layer is the one
    /// entry.
    bool is_entry_layer (int layer) const {
        return num_entry_layers() ? m_layers[layer]->entry_layer()
-                                  : (layer == nlayers()-1);
+                                  : is_last_layer(layer);
    }

 private:
@@ -42,6 +42,11 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 %option noyywrap
 %option nounput noinput

+/* never-interactive fixes a Windows compatibility problem where the
+ * lexer emits isatty calls that don't exist.
+ */
+%option never-interactive
+
 /* Option 'prefix' creates a C++ lexer with the given prefix, so that
  * we can link with other flex-generated lexers in the same application
  * without name conflicts.
@@ -106,6 +111,11 @@ using namespace OSL::pvt;
 #define YY_NO_UNISTD_H
 #endif

+#ifdef _MSC_VER
+#include <io.h>
+#define isatty _isatty
+#endif
+
 %}

 /* Declare modes */
@@ -2044,20 +2044,17 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
    for (int opnum = beginop;  opnum < endop;  opnum += 1) {
        ASSERT (old_num_ops == num_ops); // better not happen unknowingly
        DASSERT (num_ops == inst()->ops().size());
+        DASSERT (size_t(opnum) < inst()->ops().size());
        if (m_stop_optimizing)
            break;
-        // Before getting a reference to this op, be sure that a space
-        // is reserved at the end in case a folding routine inserts an
-        // op.  That ensures that the reference won't be invalid.
-        inst()->ops().reserve (num_ops+1);
-        Opcode &op (inst()->ops()[opnum]);
+        Opcode *op = &inst()->ops()[opnum];
        if (skipops) {
            // If a previous optimization inserted ops and told us
            // to skip over the new ones, we still need to unalias
            // any symbols written by this op, but otherwise skip
            // all subsequent optimizations until we run down the
            // skipops counter.
-            block_unalias_written_args (op);
+            block_unalias_written_args (*op);
            ASSERT (lastblock == m_bblockids[opnum] &&
                    "this should not be a new basic block");
            --skipops;
@@ -2071,23 +2068,23 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
            lastblock = m_bblockids[opnum];
        }
        // Nothing below here to do for no-ops, take early out.
-        if (op.opname() == u_nop)
+        if (op->opname() == u_nop)
            continue;
        // De-alias the readable args to the op and figure out if
        // there are any constants involved.
-        for (int i = 0, e = op.nargs();  i < e;  ++i) {
-            if (! op.argwrite(i)) { // Don't de-alias args that are written
-                int argindex = op.firstarg() + i;
+        for (int i = 0, e = op->nargs();  i < e;  ++i) {
+            if (! op->argwrite(i)) { // Don't de-alias args that are written
+                int argindex = op->firstarg() + i;
                int argsymindex = dealias_symbol (inst()->arg(argindex), opnum);
                inst()->args()[argindex] = argsymindex;
            }
-            if (op.argread(i))
-                use_stale_sym (oparg(op,i));
+            if (op->argread(i))
+                use_stale_sym (oparg(*op,i));
        }
        // If it's a simple assignment and the lvalue is "stale", go
        // back and eliminate its last assignment.
-        if (is_simple_assign(op))
-            simple_sym_assign (oparg (op, 0), opnum);
+        if (is_simple_assign(*op))
+            simple_sym_assign (oparg (*op, 0), opnum);
        // Make sure there's room for several more symbols, so that we
        // can add a few consts if we need to, without worrying about
        // the addresses of symbols changing when we add a new one below.
@@ -2095,7 +2092,7 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
        // For various ops that we know how to effectively
        // constant-fold, dispatch to the appropriate routine.
        if (optimize() >= 2 && m_opt_constant_fold) {
-            const OpDescriptor *opd = shadingsys().op_descriptor (op.opname());
+            const OpDescriptor *opd = shadingsys().op_descriptor (op->opname());
            if (opd && opd->folder) {
                int c = (*opd->folder) (*this, opnum);
                if (c) {
@@ -2105,23 +2102,24 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
                    skipops = num_ops - old_num_ops;
                    endop += num_ops - old_num_ops; // adjust how far we loop
                    old_num_ops = num_ops;
+                    op = &inst()->ops()[opnum];  // in case ops resized
                }
            }
        }
        // Clear local block aliases for any args that were written
        // by this op
-        block_unalias_written_args (op);
+        block_unalias_written_args (*op);

        // Now we handle assignments.
-        if (optimize() >= 2 && op.opname() == u_assign && m_opt_assign)
-            changed += optimize_assignment (op, opnum);
+        if (optimize() >= 2 && op->opname() == u_assign && m_opt_assign)
+            changed += optimize_assignment (*op, opnum);
        if (optimize() >= 2 && m_opt_elide_useless_ops)
-            changed += useless_op_elision (op, opnum);
+            changed += useless_op_elision (*op, opnum);
        if (m_stop_optimizing)
            break;
        // Peephole optimization involving pair of instructions (the second
        // instruction will be in the same basic block.
-        if (optimize() >= 2 && m_opt_peephole && op.opname() != u_nop) {
+        if (optimize() >= 2 && m_opt_peephole && op->opname() != u_nop) {
            // Find the next instruction in the same basic block
            int op2num = next_block_instruction (opnum);
            if (op2num) {
@@ -2133,6 +2131,7 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
                    // skipops = num_ops - old_num_ops;
                    endop += num_ops - old_num_ops; // adjust how far we loop
                    old_num_ops = num_ops;
+                    op = &inst()->ops()[opnum];  // in case ops resized
                }
            }
        }
@@ -2140,14 +2139,14 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
        // Special cases for "if", "functioncall", and loops: Optimize the
        // sequences of instructions in the bodies recursively in a way that
        // allows us to be clever about the basic block alias tracking.
-        ustring opname = op.opname();
+        ustring opname = op->opname();
        if ((opname == u_if || opname == u_functioncall ||
             opname == u_for || opname == u_while || opname == u_dowhile)
              && shadingsys().m_opt_seed_bblock_aliases) {
            // Find all symbols written anywhere in the instruction range
            // of the bodies.
            FastIntSet symwrites;
-            catalog_symbol_writes (opnum+1, op.farthest_jump(), symwrites);
+            catalog_symbol_writes (opnum+1, op->farthest_jump(), symwrites);
            // Save the aliases from the basic block we are exiting.
            // If & function call: save all prior aliases.
            // Loops: dont save aliases involving syms written in the loop.
@@ -2163,11 +2162,11 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
            // one (the body), loops have 4 (init, cond, body, incr),
            int njumps = (opname == u_if) ? 2 : (opname == u_functioncall ? 1 : 4);
            // Recursively optimize each body block.
-            // Don't use op after inserstions! Use inst()->op(opnum).
-            for (int j = 0; j < njumps; ++j)
-                changed += optimize_ops (j==0 ? opnum+1 : inst()->op(opnum).jump(j-1),
-                                         inst()->op(opnum).jump(j),
-                                         &saved_block_aliases);
+            for (int j = 0; j < njumps; ++j) {
+                changed += optimize_ops (j==0 ? opnum+1 : op->jump(j-1),
+                                         op->jump(j), &saved_block_aliases);
+                op = &inst()->ops()[opnum];  // in case ops resized
+            }
            // Adjust optimization loop end if any instructions were added
            num_ops = inst()->ops().size();
            endop += num_ops - old_num_ops;
@@ -2182,13 +2181,13 @@ RuntimeOptimizer::optimize_ops (int beginop, int endop,
                restored_aliases.swap (saved_block_aliases);
                // catalog again, in case optimizations in those blocks
                // caused writes that weren't apparent before.
-                catalog_symbol_writes (opnum+1, inst()->op(opnum).farthest_jump(), symwrites);
+                catalog_symbol_writes (opnum+1, op->farthest_jump(), symwrites);
                copy_block_aliases (restored_aliases, saved_block_aliases,
                                    &symwrites);
            }
            seed_block_aliases = &saved_block_aliases;
            // Get ready to increment to the next instruction
-            opnum = inst()->op(opnum).farthest_jump() - 1;
+            opnum = op->farthest_jump() - 1;
        }
    }
    m_block_aliases_stack.pop_back();  // Done with saved_block_aliases
@@ -2356,7 +2355,10 @@ RuntimeOptimizer::resolve_isconnected ()
                ASSERT (fieldsymid >= 0);
                s = inst()->symbol(fieldsymid);
            }
-            int val = (s->connected() ? 1 : 0) + (s->connected_down() ? 2 : 0);
+            bool upconnected = s->connected();
+            if (!s->lockgeom() && shadingsys().userdata_isconnected())
+                upconnected = true;
+            int val = (upconnected ? 1 : 0) + (s->connected_down() ? 2 : 0);
            turn_into_assign (op, add_constant(TypeDesc::TypeInt, &val),
                              "resolve isconnected()");
        }
@@ -631,7 +631,7 @@ ShadingSystemImpl::ShadingSystemImpl (RendererServices *renderer,
    : m_renderer(renderer), m_texturesys(texturesystem), m_err(err),
      m_statslevel (0), m_lazylayers (true),
      m_lazyglobals (true), m_lazyunconnected(true),
-      m_lazy_userdata(false),
+      m_lazy_userdata(false), m_userdata_isconnected(false),
      m_clearmemory (false), m_debugnan (false), m_debug_uninit(false),
      m_lockgeom_default (true), m_strict_messages(true),
      m_range_checking(true),
@@ -1034,6 +1034,7 @@ ShadingSystemImpl::attribute (string_view name, TypeDesc type,
    ATTR_SET ("lazyglobals", int, m_lazyglobals);
    ATTR_SET ("lazyunconnected", int, m_lazyunconnected);
    ATTR_SET ("lazy_userdata", int, m_lazy_userdata);
+    ATTR_SET ("userdata_isconnected", int, m_userdata_isconnected);
    ATTR_SET ("clearmemory", int, m_clearmemory);
    ATTR_SET ("debug_nan", int, m_debugnan);
    ATTR_SET ("debugnan", int, m_debugnan);  // back-compatible alias
@@ -1137,6 +1138,7 @@ ShadingSystemImpl::getattribute (string_view name, TypeDesc type,
    ATTR_DECODE ("lazyglobals", int, m_lazyglobals);
    ATTR_DECODE ("lazyunconnected", int, m_lazyunconnected);
    ATTR_DECODE ("lazy_userdata", int, m_lazy_userdata);
+    ATTR_DECODE ("userdata_isconnected", int, m_userdata_isconnected);
    ATTR_DECODE ("clearmemory", int, m_clearmemory);
    ATTR_DECODE ("debug_nan", int, m_debugnan);
    ATTR_DECODE ("debugnan", int, m_debugnan);  // back-compatible alias
@@ -1558,6 +1560,7 @@ ShadingSystemImpl::getstats (int level) const
    BOOLOPT (lazyglobals);
    BOOLOPT (lazyunconnected);
    BOOLOPT (lazy_userdata);
+    BOOLOPT (userdata_isconnected);
    BOOLOPT (clearmemory);
    BOOLOPT (debugnan);
    BOOLOPT (debug_uninit);
@@ -2121,7 +2124,6 @@ ShadingSystemImpl::ShaderGroupBegin (string_view groupname,
            // Zero-pad if we parsed fewer values than we needed
            intvals.resize (type.numelements()*type.aggregate, 0);
            ASSERT (int(type.numelements())*type.aggregate == int(intvals.size()));
-            Parameter (paramname, type, &intvals[0], lockgeom);
        } else if (type.basetype == TypeDesc::FLOAT) {
            floatvals.clear ();
            floatvals.reserve (vals_to_preallocate);
@@ -2141,7 +2143,6 @@ ShadingSystemImpl::ShaderGroupBegin (string_view groupname,
            // Zero-pad if we parsed fewer values than we needed
            floatvals.resize (type.numelements()*type.aggregate, 0);
            ASSERT (int(type.numelements())*type.aggregate == int(floatvals.size()));
-            Parameter (paramname, type, &floatvals[0], lockgeom);
        } else if (type.basetype == TypeDesc::STRING) {
            stringvals.clear ();
            stringvals.reserve (vals_to_preallocate);
@@ -2171,7 +2172,6 @@ ShadingSystemImpl::ShaderGroupBegin (string_view groupname,
            // Zero-pad if we parsed fewer values than we needed
            stringvals.resize (type.numelements()*type.aggregate, ustring());
            ASSERT (int(type.numelements())*type.aggregate == int(stringvals.size()));
-            Parameter (paramname, type, &stringvals[0], lockgeom);
        }

        if (Strutil::parse_prefix (p, "[[")) {  // hints
@@ -3180,5 +3180,3 @@ osl_bind_interpolated_param (void *sg_, const void *name, long long type,
    }
    return 0;  // no such user data
 }
-
-
@@ -12,7 +12,6 @@ else ()
    ADD_LIBRARY ( oslquery SHARED ${liboslquery_srcs} )
 endif ()

-TARGET_LINK_LIBRARIES ( oslquery ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} )
-LINK_ILMBASE ( oslquery )
+TARGET_LINK_LIBRARIES ( oslquery ${OPENIMAGEIO_LIBRARY} ${ILMBASE_LIBRARIES} ${Boost_LIBRARIES} )

 INSTALL ( TARGETS oslquery RUNTIME DESTINATION bin LIBRARY DESTINATION lib ARCHIVE DESTINATION lib )
@@ -8,7 +8,6 @@ if (BUILDSTATIC)
 endif ()

 ADD_EXECUTABLE ( oslc ${oslc_srcs} )
-LINK_ILMBASE ( oslc )
 TARGET_LINK_LIBRARIES ( oslc oslcomp ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
 INSTALL ( TARGETS oslc RUNTIME DESTINATION bin )

@@ -37,6 +37,7 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

 #include <OpenImageIO/filesystem.h>
 #include <OpenImageIO/sysutil.h>
+#include <OpenImageIO/thread.h>

 #include "OSL/oslcomp.h"
 #include "OSL/oslexec.h"
@@ -1,5 +1,4 @@
 SET ( oslinfo_srcs oslinfo.cpp )
 ADD_EXECUTABLE ( oslinfo ${oslinfo_srcs} )
-LINK_ILMBASE ( oslinfo )
 TARGET_LINK_LIBRARIES ( oslinfo oslquery ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
 INSTALL ( TARGETS oslinfo RUNTIME DESTINATION bin )
@@ -1,6 +1,5 @@
 # The 'testrender' executable
 FILE(GLOB testrender_src *.cpp)
 ADD_EXECUTABLE ( testrender ${testrender_src} )
-LINK_ILMBASE ( testrender )
 TARGET_LINK_LIBRARIES ( testrender oslexec oslcomp oslquery ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
 INSTALL ( TARGETS testrender RUNTIME DESTINATION bin )
@@ -37,6 +37,7 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <OpenImageIO/argparse.h>
 #include <OpenImageIO/strutil.h>
 #include <OpenImageIO/timer.h>
+#include <OpenImageIO/thread.h>

 #ifdef USE_EXTERNAL_PUGIXML
 # include <pugixml.hpp>
@@ -1,7 +1,6 @@
 # The 'testshade' executable
 SET ( testshade_srcs testshade.cpp simplerend.cpp )
 ADD_EXECUTABLE ( testshade ${testshade_srcs} testshademain.cpp )
-LINK_ILMBASE ( testshade )
 TARGET_LINK_LIBRARIES ( testshade oslexec oslcomp oslquery ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
 INSTALL ( TARGETS testshade RUNTIME DESTINATION bin )

@@ -19,8 +18,6 @@ INSTALL ( TARGETS libtestshade RUNTIME DESTINATION bin LIBRARY DESTINATION lib A

 # The 'testshade_dso' executable
 ADD_EXECUTABLE ( testshade_dso testshade_dso.cpp )
-LINK_ILMBASE ( testshade_dso )
 TARGET_LINK_LIBRARIES ( testshade_dso ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS} )
-#oslexec oslcomp oslquery ${OPENIMAGEIO_LIBRARY} ${Boost_LIBRARIES} ${CMAKE_DL_LIBS})
 INSTALL ( TARGETS testshade_dso RUNTIME DESTINATION bin )

@@ -78,6 +78,7 @@ static bool use_group_outputs = false;
 static bool do_oslquery = false;
 static bool inbuffer = false;
 static bool use_shade_image = false;
+static bool userdata_isconnected = false;
 static int xres = 1, yres = 1;
 static int num_threads = 0;
 static std::string groupname;
@@ -133,6 +134,7 @@ set_shadingsys_options ()
    shadingsys->attribute ("lockgeom", 1);
    shadingsys->attribute ("debug_nan", debugnan);
    shadingsys->attribute ("debug_uninit", debug_uninit);
+    shadingsys->attribute ("userdata_isconnected", userdata_isconnected);
    if (! shaderpath.empty())
        shadingsys->attribute ("searchpath:shader", shaderpath);
    shadingsys_options_set = true;
@@ -469,6 +471,7 @@ getargs (int argc, const char *argv[])
                "--expr %@ %s", &specify_expr, NULL, "Specify an OSL expression to evaluate",
                "--offsetst %f %f", &soffset, &toffset, "Offset s & t texture coordinates (default: 0 0)",
                "--scalest %f %f", &sscale, &tscale, "Scale s & t texture lookups (default: 1, 1)",
+                "--userdata_isconnected", &userdata_isconnected, "Consider lockgeom=0 to be isconnected()",
                "-v", &verbose, "Verbose output",
                NULL);
    if (ap.parse(argc, argv) < 0 || (shadernames.empty() && groupspec.empty())) {
@@ -0,0 +1,3 @@
+Compiled test.osl -> test.oso
+a = 26 (hex 1a), b = -1 (hex ffffffff)
+
@@ -0,0 +1,3 @@
+#!/usr/bin/env python
+
+command = testshade("test")
@@ -0,0 +1,7 @@
+shader
+test ()
+{
+    int a = 0x001a;
+    int b = 0xffffffff;
+    printf ("a = %d (hex %x), b = %d (hex %x)\n", a, a, b, b);
+}
@@ -13,6 +13,26 @@ a=0.5
 Downstream:
 a connected: 1  (value=0.5)
 b connected: 0  (value=0)
+c connected: 0  (value=0)
+mystruct1 connected: 1  (value=0)
+mystruct1.x connected: 1  (value=3)
+mystruct2 connected: 0  (value=3)
+mystruct2.x connected: 0  (value=0)
+
+Connect upstream.out to downstream.a
+Connect upstream.struct1 to downstream.mystruct1
+Upstream:
+out connected: 2  (value=0.5)
+notout connected: 0  (value=10)
+struct1 connected: 2  (value=10)
+struct1.x connected: 2  (value=3)
+struct2 connected: 0  (value=3)
+struct2.x connected: 0  (value=5)
+a=0.5
+Downstream:
+a connected: 1  (value=0.5)
+b connected: 0  (value=0)
+c connected: 1  (value=0)
 mystruct1 connected: 1  (value=0)
 mystruct1.x connected: 1  (value=3)
 mystruct2 connected: 0  (value=3)
@@ -1,5 +1,12 @@
 #!/usr/bin/env python

-command = testshade("-g 1 1 --layer upstream upstream --layer downstream test " +
+command  = testshade("--layer upstream upstream --layer downstream test " +
+                     "--connect upstream out downstream a " +
+                     "--connect upstream struct1 downstream mystruct1")
+
+# Run again, this time considering userdata (lockgeom=0) parameters to be
+# isconnected(). This time, parameter c should look like a connection.
+command += testshade("--userdata_isconnected " +
+                     "--layer upstream upstream --layer downstream test " +
                     "--connect upstream out downstream a " +
                     "--connect upstream struct1 downstream mystruct1")
@@ -16,6 +16,7 @@ void status (MyStruct variable, string name)


 shader test (float a = 0, float b = 0,
+             float c = 0 [[ int lockgeom=0 ]],
             MyStruct mystruct1 = {0,0},
             MyStruct mystruct2 = {0,0})
 {
@@ -23,6 +24,7 @@ shader test (float a = 0, float b = 0,
    printf ("Downstream:\n");
    status (a, "a");
    status (b, "b");
+    status (c, "c");
    status (mystruct1, "mystruct1");
    status (mystruct2, "mystruct2");
 }
@@ -3,6 +3,7 @@ Parameter initialization:
  matrixparam0  = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrixparam1  = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
  matrixparam1m = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrixparam1world = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

  matrix (0.1) = 0.1 0 0 0 0 0.1 0 0 0 0 0.1 0 0 0 0 0.1
  matrix (0.1, 0.2, 0.3, 0.4,  0.5, 0.6, 0.7, 0.8,  0.9, 1, 1.1, 1.2,  1.3, 1.4, 1.5, 1.6) 
@@ -20,6 +21,8 @@ testing with spaces:
  matrix ("shader", 0) = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrix ("shader", 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1) 
 	= 0 0.70711 0 0 0 0.70711 0 0 0 0 1 0 0 0 0 1
+  matrix ("world", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrix ("common", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

 Testing matrix from-to construction:
  matrix ("common", "shader") = (0.70711 -0.70711 0 0 0.70711 0.70711 0 0 0 0 1 0 -0.70711 0.70711 0 1)
@@ -65,6 +68,7 @@ Parameter initialization:
  matrixparam0  = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrixparam1  = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
  matrixparam1m = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrixparam1world = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

  matrix (0.1) = 0.1 0 0 0 0 0.1 0 0 0 0 0.1 0 0 0 0 0.1
  matrix (0.1, 0.2, 0.3, 0.4,  0.5, 0.6, 0.7, 0.8,  0.9, 1, 1.1, 1.2,  1.3, 1.4, 1.5, 1.6) 
@@ -82,6 +86,8 @@ testing with spaces:
  matrix ("shader", 1) = 0.70711 0.70711 0 0 -0.70711 0.70711 0 0 0 0 1 0 1 0 0 1
  matrix ("shader", 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1) 
 	= 0.70711 0.70711 0 0 -0.70711 0.70711 0 0 0 0 1 0 1 0 0 1
+  matrix ("world", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrix ("common", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

 Testing matrix from-to construction:
  matrix ("common", "shader") = (0.70711 -0.70711 0 0 0.70711 0.70711 0 0 0 0 1 0 -0.70711 0.70711 0 1)
@@ -126,6 +132,7 @@ Parameter initialization:
  matrixparam0  = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrixparam1  = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
  matrixparam1m = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrixparam1world = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

  matrix (0.1) = 0.1 0 0 0 0 0.1 0 0 0 0 0.1 0 0 0 0 0.1
  matrix (0.1, 0.2, 0.3, 0.4,  0.5, 0.6, 0.7, 0.8,  0.9, 1, 1.1, 1.2,  1.3, 1.4, 1.5, 1.6) 
@@ -143,6 +150,8 @@ testing with spaces:
  matrix ("shader", 0) = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrix ("shader", 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1) 
 	= 0 0.70711 0 0 0 0.70711 0 0 0 0 1 0 0 0 0 1
+  matrix ("world", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrix ("common", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

 Testing matrix from-to construction:
  matrix ("common", "shader") = (0.70711 -0.70711 0 0 0.70711 0.70711 0 0 0 0 1 0 -0.70711 0.70711 0 1)
@@ -187,6 +196,7 @@ Parameter initialization:
  matrixparam0  = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  matrixparam1  = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
  matrixparam1m = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrixparam1world = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

  matrix (0.1) = 0.1 0 0 0 0 0.1 0 0 0 0 0.1 0 0 0 0 0.1
  matrix (0.1, 0.2, 0.3, 0.4,  0.5, 0.6, 0.7, 0.8,  0.9, 1, 1.1, 1.2,  1.3, 1.4, 1.5, 1.6) 
@@ -204,6 +214,8 @@ testing with spaces:
  matrix ("shader", 1) = 0.70711 0.70711 0 0 -0.70711 0.70711 0 0 0 0 1 0 1 0 0 1
  matrix ("shader", 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1) 
 	= 0.70711 0.70711 0 0 -0.70711 0.70711 0 0 0 0 1 0 1 0 0 1
+  matrix ("world", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
+  matrix ("common", 1) = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1

 Testing matrix from-to construction:
  matrix ("common", "shader") = (0.70711 -0.70711 0 0 0.70711 0.70711 0 0 0 0 1 0 -0.70711 0.70711 0 1)
@@ -23,13 +23,15 @@ void matrix_fromto (string from, string to)
 shader
 test (matrix matrixparam0 = 0,
      matrix matrixparam1 = 1,
-      matrix matrixparam1m = matrix(1))
+      matrix matrixparam1m = matrix(1),
+      matrix matrixparam1world = matrix("world",1))
 {
    // Test parameter initialization
    printf ("Parameter initialization:\n");
    printf ("  matrixparam0  = %g\n", matrixparam0);
    printf ("  matrixparam1  = %g\n", matrixparam1);
    printf ("  matrixparam1m = %g\n", matrixparam1m);
+    printf ("  matrixparam1world = %g\n", matrixparam1world);
    printf ("\n");

    // Test matrix constructors
@@ -68,6 +70,14 @@ test (matrix matrixparam0 = 0,
                a, pretty(matrix ("shader", a, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1)));
    }

+    // World and comomon space construction
+    {
+        matrix Mw = matrix ("world", 1);
+        matrix Mc = matrix ("common", 1);
+        printf ("  matrix (\"world\", 1) = %g\n", Mw);
+        printf ("  matrix (\"common\", 1) = %g\n", Mc);
+    }
+
    // Test "from-to" matrix construction and getmatrix calls
    {
        printf ("\nTesting matrix from-to construction:\n");
@@ -0,0 +1,5 @@
+shader arrayparam(
+	color in_value[] = { color(1) },
+	output color out_value = in_value[0] )
+{
+}
@@ -0,0 +1,4 @@
+Compiled arrayparam.osl -> arrayparam.oso
+Compiled simple.osl -> simple.oso
+1.000000 0.000000 1.000000
+
@@ -0,0 +1,7 @@
+#!/usr/bin/env python
+
+# This is a regression test for a bug in which instance merging was
+# done incorrectly for shaders that differed only in the values given
+# to parameters which were arrays of unspecified length.
+
+command = testshade("--group data/shadergroup")
@@ -0,0 +1,7 @@
+param color[1] in_value 0 0 1 ;
+shader arrayparam param_node_2 ;
+param color[1] in_value 1 0 0 ;
+shader arrayparam param_node_1 ;
+shader simple simple_surface ;
+connect param_node_1.out_value simple_surface.a ;
+connect param_node_2.out_value simple_surface.b ;
@@ -0,0 +1,7 @@
+surface simple(
+	color a = color(0),
+	color b = color(0) )
+{
+	printf("%f\n", a + b);
+	Ci = emission() * (a + b);
+}
@@ -0,0 +1,3 @@
+test.osl:2: error: 'cos' is not a function
+test.osl:2: error: No matching function call to 'cos (int)'
+FAILED test.osl
@@ -0,0 +1,5 @@
+#!/usr/bin/env python
+
+# command = oslc("test.osl")
+# don't even need that -- it's automatic
+failureok = 1     # this test is expected to have oslc errors
@@ -0,0 +1,9 @@
+void mybug (float cos) {
+    cos(1);  // inner scope cos is not a function
+}
+
+
+shader test ()
+{
+}
+
@@ -0,0 +1,3 @@
+test.osl:3: warning: Comma operator is very confusing here. Did you mean to use a constructor: x = vector(...)?
+test.osl:4: warning: Comma operator is very confusing here. Did you mean to use a constructor: = color(...)?
+Compiled test.osl -> test.oso
@@ -0,0 +1,4 @@
+#!/usr/bin/env python
+
+# command = oslc("test.osl")
+# don't even need that -- it's automatic
@@ -0,0 +1,5 @@
+shader test ( output color Cout = 0 )
+{
+    vector x = (1, 2, 3);
+    Cout = (u, v, 0);
+}
@@ -0,0 +1,6 @@
+Compiled test.osl -> test.oso
+shader "test"
+    "foo" "color"
+		Default value: [ 0.1 0.2 0.3 ]
+    "bar" "color"
+		Default value: [ 0.75 0.75 0.75 ]
@@ -0,0 +1,3 @@
+#!/usr/bin/env python 
+
+command = oslinfo("-v test")
@@ -0,0 +1,5 @@
+shader test (color foo=color(0.1, 0.2, 0.3),
+             color bar=color(0.75))
+{
+    printf ("foo=%f, bar=%f\n", foo, bar);
+}
@@ -0,0 +1,9 @@
+Compiled test.osl -> test.oso
+
+Output Cout to color.tif
+Output Cout_dx to dx.tif
+Output Cout_dy to dy.tif
+spline in=0.492, result = 0.0262 0.52 0.47    dx = -0.0116 -0.0394 0.0594   dy = 0 0 0
+spline in=0.508, result = 0.0163 0.48 0.53    dx = -0.00841 -0.0394 0.0594   dy = 0 0 0
+spline in=0.492, result = 0.0262 0.52 0.47    dx = -0.0116 -0.0394 0.0594   dy = 0 0 0
+spline in=0.508, result = 0.0163 0.48 0.53    dx = -0.00841 -0.0394 0.0594   dy = 0 0 0
@@ -0,0 +1,5 @@
+#!/usr/bin/env python
+
+command += testshade("-g 64 64 -od uint8 -o Cout color.tif -o Cout_dx dx.tif -o Cout_dy dy.tif test")
+
+outputs = [ "out.txt", "color.tif", "dx.tif", "dy.tif" ]
@@ -0,0 +1,27 @@
+shader test
+  (
+    float in = u,
+    string basis = "bspline",
+    int knots = 8,
+    color values[8] = {
+                        color(0,0,0),
+                        color(0,0,0),
+                        color(1,0,0),
+                        color(0,1,0),
+                        color(0,0,1),
+                        color(0,1,1),
+                        color(1,1,1),
+                        color(1,1,1)
+                      },
+    output color Cout = 0,
+    output color Cout_dx = 0,
+    output color Cout_dy = 0,
+  )
+{
+    Cout = spline (basis, in, knots, values);
+    if (u > 0.49 && u < 0.51 && v > 0.49 && v < 0.51)
+        printf ("spline in=%.3g, result = %.3g    dx = %.3g   dy = %.3g\n",
+                in, Cout, Dx(Cout), Dy(Cout));
+    Cout_dx = 0.5 + 10 * Dx(Cout);
+    Cout_dy = 0.5 + 10 * Dy(Cout);
+}
@@ -0,0 +1,4 @@
+shader attribs (output float value = 0)
+{
+    value = v * 10;
+}
@@ -0,0 +1,6 @@
+Compiled attribs.osl -> attribs.oso
+Compiled test.osl -> test.oso
+Connect A.value to root.swidth
+Connect A.value to root.twidth
+
+Output Cout to out.tif
@@ -0,0 +1,7 @@
+#!/usr/bin/env python
+
+command += testshade("-g 128 128 -layer A attribs -param:type=string filename " +
+                     "../common/textures/grid.tx -layer root test -connect A value " +
+                     "root swidth -connect A value root twidth -od uint8 -o Cout out.tif")
+outputs = [ "out.txt", "out.tif" ]
+
@@ -0,0 +1,7 @@
+surface test (string filename = "",
+              float s = u, float t = v,
+              float swidth = 1.0, float twidth = 1.0,
+              output color Cout = 0 )
+{
+    Cout = texture (filename, s, t, "swidth", swidth, "twidth", twidth);
+}
Author	SHA1	Message	Date
Larry Gritz	2b1a7ec6c8	CHANGES	2017-03-01 18:24:01 -08:00
Larry Gritz	fe42ddb15d	Minor fixes for C++11 and use of std::atomic	2016-11-07 22:42:11 -08:00
Larry Gritz	7085e06d77	Travis: change to precise linux, we don't trust trusty	2016-11-07 22:42:11 -08:00
Larry Gritz	7eb31d02c7	Post-release version bump	2016-10-31 15:48:19 -07:00
Larry Gritz	39154a5598	CHANGES	2016-10-31 15:48:19 -07:00
Larry Gritz	5dd06f031e	Typo bug transpose() vs transposed(). (#662 ) Imath::M44f.transpose() transposes in place and returns the ref, whereas transposed() returns the transpose without altering the original. Not a symptomatic bug because we only did this on the result of an inverse(), which was a temporary, so didn't hurt to mess with it in place. But better to correct this code to properly reflect our intent.	2016-10-31 15:48:19 -07:00
Larry Gritz	f7b869d89a	Adjust Travis now that OIIO master is C++11 only Still test against release OIIO with C++03	2016-10-14 18:47:37 -07:00
Larry Gritz	9680823437	Be sure to propagate C++ standard flags to the compilation of llvm_ops.cpp (#674 ) We weren't properly transferring --std=c++11 from the directives to compile OSL's implementation, to the command line that compiles the llvm bitcode from llvm_ops.cpp. They should match.	2016-10-13 10:20:38 -07:00
Larry Gritz	3e62b5343c	oslc better err reporting for use of non-function symbol as function. (#684 ) See the new testsuite example. This crashed oslc before! Fix it up to not crash (by marking the unfound function's symbol as NULL), and also improve the resulting error message. Fixes #681	2016-10-13 09:42:38 -07:00
Larry Gritz	69d213b0f1	Suppress clang 3.9 warnings (#683 )	2016-10-13 09:33:49 -07:00
Larry Gritz	16afdcbadf	Fix warning from last checkin	2016-10-04 11:12:11 -07:00
Larry Gritz	307abe1149	Bug fix: matrix param initialized with matrix(space,1) left out init ops (#675 ) When generating the static construction values, it needed to recognize that case and set the "completed" flag to false, so that it would know that it couldn't be constructed statically and needed init ops emitted for that paramter.	2016-10-04 11:11:58 -07:00
Larry Gritz	08e87f05b0	Bug fix: incorrect dropping of optional texture arguments. There was a well-meaning optimization: dropping the setting of TextureOpt fields when the value was known to overwrite the field with the default value. But of course this should only happen when the value is CONSTANT, it shouldn't happen just because the value passed is a param and the parameter's default value is the option default value -- because the param may be connected to an earlier layer that passes a different value. The test is fixed now, the optimization only kicks in when the value passed is definitely a constant. (Unconnected params should have been turned into consts before this code is encountered, so it should still provide the optimization all the places where it was originally intended to work.) Fixes #670.	2016-09-27 15:45:25 -07:00
Larry Gritz	2d704ce671	Fix variant of spline() that didn't compute derivatives properly. (#666 ) Specifically, the case where the knots are colors (without derivs), the abscissa float with derivs.	2016-08-23 09:41:35 -07:00
Brecht Van Lommel	ab65ebe086	Fix OSL compiler assert when indexing into closure type. (#663 )	2016-08-04 16:30:31 -07:00
Larry Gritz	bacb5c53b4	post-release version bump	2016-08-01 10:11:42 -07:00
Larry Gritz	7dc3928cf0	CHANGES	2016-08-01 10:09:59 -07:00
Larry Gritz	f7cc4e595f	oslc didn't understand hex integer constants (like 0x01fc). #653	2016-07-06 17:15:21 -07:00
Larry Gritz	3a882f40b9	Bug fix: incorrect instance merging when unspecified-length arrays differed. (#656 ) Because the altered data offset wasn't transferred from the instance overrides to the instance's full symbol table, it was comparing the default values rather than the actual values, so two instances that should not have been merged, were merged. Also fixed a (harmless) bug in ShadingSystemImpl::ShaderGroupBegin, where the serialized group is parsed, where it was double-calling Parameter() to add the parameters. Fixes #655	2016-07-06 17:13:36 -07:00
Larry Gritz	cc3a0f3a53	Post-release version bump	2016-07-01 17:10:50 -07:00
Larry Gritz	b1d621519d	Update PDF docs	2016-07-01 17:10:15 -07:00
Larry Gritz	c9d7f7d5f5	CHANGES	2016-07-01 17:10:15 -07:00
Larry Gritz	af8420ff2a	Fix subtle bug in optimize_ops that could seg fault. (#651 ) The optimize_ops method had a convenience reference to the op being examined in the loop. We carefully do a reserve() first to make sure that there was enough room for one insertion so that the reference would still be valid if the constant-folding function has to add an op. But this relied on the assumption that no folder would insert more than one op as it did its work, which was always true (once). Oops, much later, a new way to speed up "mix" ops had a particular corner case in which up to 7 new ops might be inserted, and I never went back and fixed the now-broken assumption here. The fix here is to change the reference to the op (which cannot be changed dynamically) into a pointer to the op, which can be re-pointed after calling the folding function to make sure that it's valid and pointing to the (potentially new) correct position as it continues on in the loop. So it is no longer sensitive to changes to the vector allocation that might happen inside the folder. I haven't changed this code for a long time. It's shocking that no problems cropped up from this bug until this week! I guess it's not a very common "corner case".	2016-07-01 17:10:15 -07:00
Larry Gritz	f4cc52f4be	Include file fixes after recent OIIO change that split atomic.h & thread.h (#650 )	2016-07-01 16:42:20 -07:00
Larry Gritz	5c4708b34f	Add gcc6 to the test matrix and fix resulting warnings.	2016-06-22 12:48:27 -07:00
Larry Gritz	4401bcba26	Make sure spec documents #pragma once	2016-06-14 11:52:53 -07:00
Larry Gritz	eff0535bc0	Windows fix for lexer (#646 ) Option "never-interactive" prevents flex from emitting calls to isatty, which is not always present on Windows. This is fine because we never use the lexer in interactive mode anyway.	2016-06-13 14:25:43 -07:00
Elvic Liang	3c7e081bb4	Fixed build break on Windows. (643)	2016-06-10 13:33:58 -07:00
Larry Gritz	f93f04c823	Don't test or set "layer ran bits" on last layer. Also be sure that the entry point doesn't use the "fast" calling convention.	2016-06-08 10:04:08 -07:00
Larry Gritz	882d6211e7	Improve docs: * Explain rules for connections * Explain that getattribute retrieves userdata	2016-05-01 08:57:03 -07:00
Larry Gritz	eb483e08a0	Fix oslc incorrectly emitting init ops for constructor color(float). We're talking about a construct like: shader foo (color c = color(1), ...) It got it right if it was color c = color(1,1,1) In both cases, it properly reduced the default value to a static constant value, but the bug was that it ALSO (unnecessarily) had extra init ops for that parameter. That would be eliminated in runtime optimization, the shader certainly ran correctly, but it threw off oslinfo, which figures that if init ops are present, it can't figure out what the default values were.	2016-04-14 10:46:10 -07:00
Stefan Büttner	2122621386	Restrict find_program llvm to given path LLVM can be installed on a system without clang. So if LLVM_DIRECTORY pointed to a locally installed llvm/clang version but the system had llvm without clang installed, until now the script would find the system LLVM without clang and the build would fail. And since LLVM is globally searched for if the hinted version failed, it should not hurt to only search in the hinted paths.	2016-04-11 17:10:17 -07:00
Larry Gritz	d132ce65d9	Warn for common misstep of forgetting constructor name. (635) shader test ( output color Cout = 0 ) { vector x = (1, 2, 3); } See where this goes wrong? This code will make x be (3,3,3). If you don't know why, think about how the comma operator works in C, C++, and related languages. The user almost certainly meant vector x = vector (1, 2, 3); With this change, the original (buggy) code will get the following warning output: test.osl:3: warning: Comma operator is very confusing here. Did you mean to use a constructor: x = vector(...)?	2016-04-05 16:23:47 -07:00
Larry Gritz	e30aeb3c69	Make an option to have OSL isconnected() return 1 for userdata parameters. This adds a new ShadingSystem attribute, "userdata_isconnected". The default of 0 is the old/usual behavior that most applications want. However, setting this attribute to 1 will cause the OSL built-in function isconnected(paramname) to return 1 not only for connected parameters, but also for parameters that may be bound to userdata (i.e., those that have lockgeom=0). The use case for this involves a renderer implementation that can mix C++ and OSL nodes within a shading group. The outputs exported by the C++ nodes propagate their values to the downstream node via get_userdata(), hijacking the mechanism that is generally used geometric primitive variable interpolation. But from the point of view of the downstream node, they should semantically appear to be a connection.	2016-04-05 16:23:47 -07:00
Larry Gritz	446d3303c1	bump release	2016-04-05 16:23:46 -07:00
Larry Gritz	e9b5e2fdb1	Fix missing -lpthread	2016-03-01 14:52:21 -08:00
Larry Gritz	99c4ece489	CHANGES	2016-02-26 21:40:03 -08:00
Larry Gritz	e8e84a5a2c	Port FindOpenEXR.cmake fixes from OIIO	2016-02-26 21:31:30 -08:00
Larry Gritz	91b5a309cb	Rewrite of FindOpenEXR.cmake	2016-02-26 21:30:36 -08:00
Larry Gritz	03724a9236	Travis: build OIIO optimized, even when testing DEBUG for OSL (for speed of CI tests).	2016-02-26 21:30:25 -08:00
Larry Gritz	2457d36432	For stdosl.h funcs marked builtin, auto-detect funny read-write cases. #620 As Chris points out in #618, the fact that params are marked as output is all we need to know that function's parameters are not all read-only (duh).	2016-02-26 21:29:49 -08:00
Larry Gritz	9a5b82f693	New ref images for recent OIIO texture improvement. Related to https://github.com/OpenImageIO/oiio/pull/1369 We can remove the old-oiio1.5 images once 1.6 is the minimum standard for OSL.	2016-02-26 14:36:10 -08:00
Larry Gritz	205c8fadd9	Fix typos in docs Fixes #622, Fixes #623	2016-02-22 21:21:13 -08:00
Larry Gritz	3e12071012	oslc bug -- getmatrix failed to note that the matrix argument is written. #618 This could lead to a cascade of optimization errors at runtime. I have no idea why this never showed up as a problem before today. In addition to fixing it on the oslc side, add a hacky fix to make sure that old compiled shaders are retroactively fixed as they are loaded at runtime.	2016-02-06 17:29:33 -08:00
Larry Gritz	86082e94a5	Better handling of multiple SIMD args (such as USE_SIMD=avx,f16c). It botched this when passing along to precompiling for LLVM.	2016-02-06 17:29:33 -08:00
Larry Gritz	de7935a7e8	Fix typos in docs. Fixes #614	2016-02-06 17:29:33 -08:00
Larry Gritz	07782fcd28	Post-branch version bump	2016-01-29 11:20:30 -08:00
Larry Gritz	02cda6aaf0	Version/release	2016-01-29 11:15:22 -08:00