CXX Std: Remember <variant> Impl. #1128
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We use `<variant>` or `<mpark/variant.hpp>` in our public API interface for datatypes, depending on the C++ standard. This pull request makes sure that the same implementation is used in downstream code, even if the C++ standard is switched. This avoids ABI issues when, e.g., using a C++14 built openPMD-api in a C++17 downstream code.
ax3l
added a commit
to ax3l/openpmd-api-feedstock
that referenced
this pull request
Oct 19, 2021
Apply patch from openPMD/openPMD-api#1128
This was referenced Oct 19, 2021
ax3l
added a commit
to ax3l/WarpX
that referenced
this pull request
Oct 19, 2021
franzpoeschel
approved these changes
Oct 19, 2021
|
Very nice, thank you for the fix :) |
RemiLehe
pushed a commit
to ECP-WarpX/WarpX
that referenced
this pull request
Oct 21, 2021
* CI: Windows + openPMD * Windows: Work-Around <variant> mismatch Build openPMD-api in C++17 as well, so we consistently use `<variant>` over `<mpark/variant.hpp>`. * Test Fix for 0.14.3 (openPMD-api 1128) Test PR in openPMD/openPMD-api#1128 * Finalize CI for now
ax3l
added a commit
to ax3l/openPMD-api
that referenced
this pull request
Nov 3, 2021
We use `<variant>` or `<mpark/variant.hpp>` in our public API interface for datatypes, depending on the C++ standard. This pull request makes sure that the same implementation is used in downstream code, even if the C++ standard is switched. This avoids ABI issues when, e.g., using a C++14 built openPMD-api in a C++17 downstream code.
ax3l
added a commit
that referenced
this pull request
Nov 4, 2021
* Read: time/dt also in long double (#1096) * Python: time/dt round-trip Test writing and reading time and dt on an iteration via properties. * Fix: Iteration read of long double time Support reading of `dt` and `time` attributes if they are of type `long double`. (openPMD standard: all `floatX` supported) * Executables: CXX_STANDARD/EXTENSIONS (#1102) Set `CXX_EXTENSIONS OFF` and `CXX_STANDARD_REQUIRED ON` for created executables. This mitigates issues with NVCC 11.0 and C++17 builds seen as added `-std=gnu++17` flags that lead to ``` nvcc fatal : Value 'gnu++17' is not defined for option 'std' ``` when using `nvcc` as CXX compiler directly. * Doc: More Locations -DPython_EXECUTABLE (#1104) Mention the `-DPython_EXECUTABLE` twice more in build examples. * NVCC + C++17 (#1103) * NVCC + C++17 Work-around a build issue with NVCC in C++17 builds. ``` include/openPMD/backend/Attributable.hpp(437): error #289: no instance of constructor "openPMD::Attribute::Attribute" matches the argument list argument types are: (std::__cxx11::string) detected during instantiation of "__nv_bool openPMD::AttributableInterface::setAttribute(const std::__cxx11::string &, T) [with T=std::__cxx11::string]" ``` from ``` inline bool AttributableInterface::setAttribute( std::string const & key, char const value[] ) { return this->setAttribute(key, std::string(value)); } ``` Seen with: - NVCC 11.0.2 + GCC 8.3.0 - NVCC 11.0.2 + GCC 7.5.0 * NVCC 11.0.2 C++17 work-around: Add Comment * Lazy parsing: Make findable in docs and use in openpmd-ls (#1111) * Use deferred iteration parsing in openpmd-ls * Make lazy/deferred parsing searchable * Add a way to search for usesteps key * HDF5: Document HDF5_USE_FILE_LOCKING (#1106) Document a HDF5 read work-around that we currently need on OLCF Jupyter (https://jupyter.olcf.ornl.gov), due to a mounting issue of GPFS in the Jupyter serice (OLCFHELP-3685). From the HDF5 1.10.1 Release Notes: ``` Other New Features and Enhancements =================================== Library ------- - Added a mechanism for disabling the SWMR file locking scheme. The file locking calls used in HDF5 1.10.0 (including patch1) will fail when the underlying file system does not support file locking or where locks have been disabled. To disable all file locking operations, an environment variable named HDF5_USE_FILE_LOCKING can be set to the five-character string 'FALSE'. This does not fundamentally change HDF5 library operation (aside from initial file open/create, SWMR is lock-free), but users will have to be more careful about opening files to avoid problematic access patterns (i.e.: multiple writers) that the file locking was designed to prevent. Additionally, the error message that is emitted when file lock operations set errno to ENOSYS (typical when file locking has been disabled) has been updated to describe the problem and potential resolution better. (DER, 2016/10/26, HDFFV-9918) ``` This also exists as a compilation option for HDF5 in CMake, where it defaults to ``TRUE`` by default, which is also what distributions/ package managers ship. Disabling from Bash: ```bash export HDF5_USE_FILE_LOCKING=FALSE ``` Disabling from Python: ```py import os os.environ['HDF5_USE_FILE_LOCKING'] = "FALSE" ``` * Avoid object slicing when deriving from Series class (#1107) * Make Series class final * Use private constructor to avoid object slicing * Doc: OMPI_MCA_io Control (#1114) Document OpenMPI MPI-I/O backend control. We have documented this long in #446. * openPMD.hpp: Include auxiliary StringManip (#1124) Include this, handy functions. * CXX Std: Remember <variant> Impl. (#1128) We use `<variant>` or `<mpark/variant.hpp>` in our public API interface for datatypes, depending on the C++ standard. This pull request makes sure that the same implementation is used in downstream code, even if the C++ standard is switched. This avoids ABI issues when, e.g., using a C++14 built openPMD-api in a C++17 downstream code. * Spack: No More `load -r` (#1125) The `-r` argument was removed from `spack load` and is now implied. * Fix AppVeyor: Python Executable (#1127) * GH Action: Add MSVC & ClangCL on Win * Fix AppVeyor: Python Executable * Avoid mismatching system Python and Conda Python * Conda: Fix Numpy * CMake: Skip Pipe Test Written in a too special way, we cannot assume SH is always present * Test 8b (Bench Read Parallel): Support Variable encoding, Fix Bugs (#1131) * added support to read variable encoding, plus fixed some bugs * fixed style * Update examples/8b_benchmark_read_parallel.cpp remove commented out code Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Update examples/8b_benchmark_read_parallel.cpp Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * removed commented line * updated 8b env option Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * HDF5 I/O optimizations (#1129) * Include HDF5 optimization options * Fix code style check * Fix validations and include checks * Fix style check * Remove unecessary strict check * Update documentation with HDF5 tuning options * Update contributions * Fix Guards for H5Pset_all_coll_metadata* * MPI Guard: H5Pset_all_coll_metadata* * Remove duplicated variable Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Include known issues section for HDF5 (#1132) * Update known issues with HDF5 and collective metadata operations * Fix rst link and tiny typo * Add targeted bugfix releases. Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> * Include check for paged allocation (#1133) * Include check for paged allocation * Update ParallelHDF5IOHandler.cpp * libfabric 1.6+: Document SST Work-Arounds (#1134) * libfabric 1.6+: Document SST Work-Arounds Document work-arounds for libfabric 1.6+ on Cray systems when using data staging / streaming with ADIOS2 SST. Co-authored-by: Franz Pöschel <franz.poeschel@gmail.com> * Fix: Read Inconsistent Zero Pads (#1118) * [Draft] Fix: Read Inconsistent Zero Pads Some codes mess up the zero-padding in `fileBased` encoding, e.g., when specifying padding to 5 digits but creating >100'000 output steps. Files like those cannot yet be parsed and fell back to no padding, which fails to open the file: ``` openpmd_00000.h5 openpmd_02000.h5 openpmd_101000.h5 openpmd_01000.h5 openpmd_100000.h5 openpmd_104000.h5 ``` Error: ``` RuntimeError: [HDF5] Failed to open HDF5 file diags/diag1/openpmd_0.h5 ``` * Revert previous changes except for test Parse iteration numbers that are longer than their padding Read inconsistent zero padding * Overflow Padding: Read Test * Warn if the prefix does end in a digit * Fix: Don't let oversize numbers accidentally bump the padding * Update test * Issue warnings on misleading patterns also when writing * Minor Style Update Co-authored-by: Franz Pöschel <franz.poeschel@gmail.com> * Release: 0.14.3 Co-authored-by: Franz Pöschel <franz.poeschel@gmail.com> Co-authored-by: guj <guj@users.noreply.github.com> Co-authored-by: Jean Luca Bez <jeanlucabez@gmail.com> Co-authored-by: Jean Luca Bez <jlbez@lbl.gov>
ax3l
added a commit
to ax3l/openPMD-api
that referenced
this pull request
Nov 5, 2021
Forgot to add in openPMD#1128, which is then missing in Python sdist files for pip from-source builds.
|
Dang, forgot to add Fix in: #1140 |
ax3l
added a commit
that referenced
this pull request
Nov 6, 2021
dpgrote
pushed a commit
to dpgrote/WarpX
that referenced
this pull request
Nov 29, 2021
* CI: Windows + openPMD * Windows: Work-Around <variant> mismatch Build openPMD-api in C++17 as well, so we consistently use `<variant>` over `<mpark/variant.hpp>`. * Test Fix for 0.14.3 (openPMD-api 1128) Test PR in openPMD/openPMD-api#1128 * Finalize CI for now
ax3l
added a commit
to ax3l/openPMD-api
that referenced
this pull request
Jan 19, 2022
Forgot to add in openPMD#1128, which is then missing in Python sdist files for pip from-source builds.
ax3l
added a commit
that referenced
this pull request
Jan 21, 2022
* Add `cmake/` to MANIFEST.in Forgot to add in #1128, which is then missing in Python sdist files for pip from-source builds. * Add failing test * Read gridSpacing also from long doubles * Python: 3.10 Add support for Python 3.10. * CI: Python 3.10 on Alpine Linux musllinux uses the musl libc on an Alpine stack, especially for docker-ish containers. Let's add coverage while also covering CPython 3.10 * Fix: Python Iteration Repr Typo Fix a typo in the `__repr__` of `Iteration` in Python. * Remove invalid records from our data structure entirely Until now, the half-parsed records were not deleted. Our sanity checks in `Series::~Series()` caught up on that and threw errors. * Fix doxygen? No idea why this PR triggers that error * Don't apply compression operators multipe times Happened so far in variable-based iteration encoding * ICC 19.1.2: CXX17 Work-Arounds (Variant) The `Attribute` constructors with implicit variant conversion sometimes do not work in ICC 19.1.2. ``` openPMD-api/src/RecordComponent.cpp(226): error: no instance of constructor "openPMD::Attribute::Attribute" matches the argument list argument types are: (openPMD::Extent) Attribute a(getExtent()); ^ openPMD-api/include/openPMD/backend/Attribute.hpp(50): note: this candidate was rejected because arguments do not match class Attribute : ^ openPMD-api/include/openPMD/backend/Attribute.hpp(50): note: this candidate was rejected because arguments do not match class Attribute : ^ openPMD-api/include/openPMD/backend/Attribute.hpp(79): note: this candidate was rejected because arguments do not match Attribute(resource r) : Variant(std::move(r)) ^ ``` Same work-around as for NVCC in #1103 * CMake 3.22+: Policy CMP0127 Fix a warning with CMake 3.22+. We use simple syntax in cmake_dependent_option, so we are compatible with the extended syntax in CMake 3.22+: https://cmake.org/cmake/help/v3.22/policy/CMP0127.html * Docs: Minor Warnings Fix minor warnings in the docs: - double reference (make 2nd ref annonymous) - misaligned reference * Docs: Fix HDF5 JSON File Fix misnamed filename and misnamed prefix in text. * Docs: Update Citation & Add BibTeX Add `.bib` blocks for easier citation of standard and API. * Docs: Fix CLI Highlighting Fix a highlighting issue for `openPMD-ls` in the docs. Explicitly use `python3` for module calls. * CMake: Warning Flags First in CXXFLAGS If we append them, then we overwrite flags like `-Wno-...` from environment variables. * ADIOS2 fixes: Incompatibilites w/ `master` Referes to post-`2.7.1` `master` branch of ADIOS2. Catch new ADIOS2 error type: string getCompressionOperator Dirtily catch ADIOS2 exception in automatically_deactivate_span Cleanly destruct a Series that errored during flushing Problem was that IO tasks that do not return cleanly but throw an exception must be removed from the IO queue, otherwise the task will be tried again during destruction. Usually, this means that anything will happen. SerialIOTest: Remove Comments (resolved) Print warning when removing task from IO queue std::string and fix warning * Automatically deactivate span-based API when operators are present * Add .pre-commit-config.yaml * Release: 0.14.4 Co-authored-by: Franz Pöschel <franz.poeschel@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We use
<variant>or<mpark/variant.hpp>in our public API interface for datatypes, depending on the C++ standard.This pull request makes sure that the same implementation is used in downstream code, even if the C++ standard is switched. This avoids ABI issues when, e.g., using a C++14 built openPMD-api in a C++17 downstream code.
Fix #1121
X-ref: