AAdrián Insaurralde AvalosSet BUILD_DYNAMIC_LINK_PYTHON to OFF, except ON for MSVC
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
Set BUILD_DYNAMIC_LINK_PYTHON to OFF, except ON for MSVC Fixes #25098 on macOS and Linux. I searched the web a bit and didn't find a better way than this if-else block. We can't use (NOT MSVC) as the value of the option, cmake syntax doesn't allow it. We can't change the value of an option() after it's first been set, so an option(... ON) in the if(MSVC) block below would do nothing. We could use SET(... ON CACHE BOOL ... FORCE) in the if(MSVC) block below, but that's a bit tricky and might be unclear. The copy-and-paste code in the if-else block is a bit ugly, but it's simple and clear. | 11 天前 | |
Clean ConfigureChecks.cmake Clean cmake | 2 年前 | |
clean FindCoin3DDoc.cmake clean FindCoin3DDoc.cmake | 3 年前 | |
CMake: Fix Eigen3 version search | 1 年前 | |
CMake: handle Arch HDF5 packaging Changes find_package() style back from config-then-legacy to legacy first. | 1 个月前 | |
Remove superflous whitespace + trailing newlines | 1 年前 | |
Build: cmake: fixes #26247 update cmake to work with new required dep lark for BIM (#26407) | 9 个月前 | |
CMake: rework HDF5 & MEDFile find_package logic Get HDF5 logic out of SetupSalomeSMESH.cmake and into its own FindHDF5.cmake, and move MEDFile related logic (OpenMPI) into FindMEDFile.cmake. Changes those two latter file to prefer loading their namesake modules straight from CMake/system, only falling back to finding those themselves when neither xyzConfig.cmake or FindXyz.cmake are available. In the case of FindHDF5.cmake, applies the _FORTIFY_SOURCE cleanup in all cases. | 1 个月前 | |
CMake: Modernize build files using outdated Python variables. | 1 年前 | |
CMake: only find Netgen when using internal smesh | 9 个月前 | |
Build: Add OCCT dev version to CMake output | 4 个月前 | |
Allow external PyCXX to be used, and find this external PyCXX via pkgconfig. | 6 个月前 | |
Build: require PySide6 QtSvgWidgets for GUI builds | 4 个月前 | |
Reduce Python rc file size * Compile Qt resources using `rcc`'s binary output mode and generate custom loading code calling `QResource.registerResource()`. This runs with the implicit options `rcc` decides to use when called with `--binary`. * Reduce FC install size as non-printable resource pack bytes were expanded to 4 bytes (`\xNN`) in the generated files. `fd --no-ignore '_rc(\.py|\.rcc)' | xargs du -bc` in the `build/Mod` dir reports a reduction from 71.0 MiB to 9.63 MiB, a 61.37 MiB reduction and 7.4× improvement. * Expand gitignore `OpenSCAD_rc.py` rule to all files generated by `PYSIDE_WRAP_RC()` because 1. none of them should be committed and 2. I banged my head for 5 minutes not understanding why `rg` was ignoring OpenSCAD specifically (I did not use initially use `--no-ignore`). * Increases performance as loading data through Python sources is slow due to the parser, data copies, etc. Measured with n = 5. Jitter can increase due to additional filesystem accesses being needed to load the data, and the speedup ratio depends on the data contained in the resources and if they were compressed by `rcc`. | Statement | Before (ms) | After (ms) | Speedup | | ---------------- | ----------------- | --------------- | ------- | | import Arch_rc | µ=261.334 σ=2.478 | µ=0.490 σ=0.029 | 533.3× | | import Draft_rc | µ=170.509 σ=1.274 | µ=1.129 σ=0.052 | 151.0× | | import Help_rc | µ= 3.785 σ=0.038 | µ=0.444 σ=0.013 | 8.5× | | import Tux_rc | µ= 2.367 σ=0.072 | µ=0.410 σ=0.033 | 5.8× | * Change all `PYSIDE_WRAP_RC()`-generated files from `SOURCE` to `DEPENDS` in `ADD_CUSTOM_TARGET()` calls: these are generated and should not be shown in IDEs, a thing `SOURCE` implies but `DEPENDS` does not. Otherwise they're equivalent from a dependency tracking perspective. * Remove the Linux differences in the `PYSIDE_WRAP_RC()` macro, as these got effectively zeroed out in 76ad87d4642e2e07e1fcbd35bf53e1ac0acbbe5a. The only remaining command difference was `PY_ATTRIBUTE` which was removed (and therefore always empty) with the drop of Qt 4 in 13b9127eeea0f1c6e8e338381110faf311efc2de. * Remove Python `Assembly_rc`. Assembly resources were compiled in and loaded twice ever since Assembly got mainlined in b7a6558c72e733ee850fedd6c6f63a87771e936a. The C++ version is more efficient and is guaranteed to be loaded (through Qt's generated static initializers) before the Python-side `import AssemblyGui` returns; and there is only one Qt resource context per process. | 3 个月前 | |
Fix whitespace issues | 3 年前 | |
Build: fix shiboken include dir when found from pip package Shiboken 6.11 on Ubuntu changed the organzation of the package, and headers are now in a shiboken6 directory in the root shiboken include dir. | 5 个月前 | |
Spelling error in variable name, caused the setting to never be applied: SPANV_LIBRARY vs SPNAV_LIBRARY | 7 年前 | |
fixes #32526 fix cmake to find py runtime deps and properly report finding ifcopenshell (#32554) * fixes #32526 fix cmake to find py runtime deps and properly report finding ifcopenshell * fixes #32526 just rely on using __version__, and not both __version__ and version * per @Lgt2x 's request add comment explaining rm of find_pip_package(${PIP_NAME}) * cmake: fix ifcopenshell discovery, rm code comments, fix spelling error * CMake find_python_runtime_dep: remove extra logging --------- Co-authored-by: Louis Gombert <louisgombert@gmail.com> | 20 天前 | |
Build: Make sure that CMake can find executables in the LibPack's scripts directory | 3 个月前 | |
Build: Drop Qt5 compatibility (#29470) * Remove Qt5 compatibility Qt5 was deprecated following the 1.1 release, this removes compatibility and forces using Qt6 to build the app. | 1 个月前 |