alembic-1.8.4-bp155.1.6

- update to 1.8.4:
  * Add ALEMBIC_DEBUG_WARNINGS_AS_ERRORS to control the build in debug mode.
  * It is set to ON by default to preserve the current behavior.
  * Use add_compile_options instead of the older add_definitions. (PR371)
  * Fix the include dirs for AbcCoreHDF5 tests. (PR382)
  * Fix issue hinted at in Issue 223: use PROJECT_SOURCE_DIR instead of
  * CMAKE_SOURCE_DIR when embedding Alembic within another CMake project. (PR393)
  * On Windows _BitScanForward64 is only available when _WIN64 is defined.
  * When it is not, fall back on using _BitScanForward. (PR 373)
  * Fuzzer issue 49213 turned up some new malformed MetaData memory problems.
  * Indexed MetaData on the write side only allows 254 (plus the default empty one)
  * small (up to 256 bytes), we now throw if we encounter malformed MetaData.
  * We were also allowing empty property and object names through on the read side
  * even though the writer forbids this.
  * Finally integrate a couple of tests flagged by Taotao Gu's local fuzzer.
  * (PR391 and PR397)
  * Fuzzer issue 52939 showed that with certain malformed MetaData the pair
  * separator should NOT occur before the assign separator. (PR396)
  * N2f and N2d properties were not being correctly bound.(PR390)

- Ogawa only supports little endian. Disable s390x builds.

- update to 1.8.3:
  * Make sure we export the IlmBase requirement in CMake when using OpenEXR 2 to
    fulfill the Imath requirement.
  * Address report from GCC Undefined Behavior Sanitizer by initializing a couple
    of bools in OXform.
  * Fix compile issue when using C++17 with Microsoft Visual Studio and use a
    hopefully much faster find first bit set when using that compiler. (Issue 358)
  * One of the fuzzer tests were was failing with 32 bit builds because of a
    different exception message.
  * Fix a bug with 32 bit memory mapped Windows builds.
  * Look for boost python a little differently, falling back on existing behavior.

- Update to 1.8.1:
  * Ogawa: Sanity check that we have a valid group to avoid
    infinite recursion during traversal.
- Enable testsuite
- Add fuzztest.patch

- Update to 1.8.0:
  API:
  * Dropped support for the ancient TR1, expect compilers to support
    at least support C++std11 and currently default to C++std14.
  * Removed ALEMBIC_LIB_USES_TR1 and ALEMBIC_LIB_USES_BOOST and
    instead expect the std versions like shared_ptr.
  Ogawa:
  * Better support emscripten and ARM toolchains by using memcpy
    to avoid dereferencing non-aligned pointers.
  AbcStitcher:
  * Propagate MetaData from the first of our ordered input archives
    over to the stitched output.
  Other:
  * Overhaul to support Imath 3
  * Fix for using Alembic as a 3rd party with Imath 3.

- update to 1.7.16:
  GitHub:
  * Add a Github Action to create a python 2.7 wheel. (PR 310)
  AbcCoreHDF5:
  * Specify char signedness when setting the HDF5 data type as some architectures,
    like ARM, default to unsigned. (PR 316)
  AbcCoreOgawa:
  * Fuzzer issue 25695: Fix leak that can occur when reading meta data, but the file
    is maliciously cut short.  (PR 315)
  * Fuzzer issue 26125:  Add a check and throw an excpetion for an obviously
    recursive group. (PR 321)

- update to 1.7.15:
  * Fuzzer issues 25166, 25175, 25185, 25204, and 25236 all test for illegal
  meta data index values which we more carefully test.
  * Fuzzer issue 25192 turned up a leak from a malformed file causing OrData to
  throw.

- update to 1.7.14:
  * Fix another Fuzzer issue with Ogawa::IData allocating too much memory
  because of a malicious file.  Add tests for Fuzzer issues:
  24846, 24853, 24598, 25051, and 25081. (PR296)
  * Remove Murmur3 from the LICENSE.txt since it was moved into the public domain
  a while ago, and is reflected as such in the code. (Issue 292)
  * Fix uninitialized member variable in OTypedGeomParam. (PR290)
  * Remove some unused local functions. (Issue 293)

- update to 1.7.13:
  * Ogawa:  Google generously setup AutoFuzz and it found several issues when
  reading maliciously created files.  These issues reported issues are fixed:
  253-258, 269-272, 282, 283 (PR266)
  * AbcImport: Add support for connecting MFnData::kFloatArray.
  * Fix issue where WFObjConvert headers were being put in the wrong place when
  installed. (Issue 264)
  * Partly address Issue 265 by making some adjustments to the Doxyfile and remove
  a very old reference to MD5Hash in the very old Houdini SOP.

- Update to version 1.7.12
  API:
  + Use critical section instead of mutex on Windows. (PR220)
  * ErrorHandler now uses const char * for message instead of
    std::string for efficiency gain. (PR221)
  + Fix incompatibilities with C++17. (PR237)
  * Add optional DCC FPS hint when calling CreateArchiveWithInfo
    and GetArchiveInfo. (PR241)
  Maya:
  + AbcExport: Support writing animated facesets. (PR217)
  + AbcExport: Support per curve width. (PR239)
  * AbcExport: When writing sparse data like just the UVs, create
    an OObject instead of an OXform. (PR241)
  Utils:
  * WFOBJConvert: Fix bug where OBJ files with more than one
    shape would fail to properly convert UVs and Normals. (PR232)
  CMake:
  * Make it easier to select the python version to build against.
    (PR215)
- Clean up spec, remove any superflous parameterization

- Avoid forcing SSE/SSE2 on %ix86, it is not guaranteed to exist.

- Move documentation to -dev subpackage.

- version 1.7.11: initial build