* [corrade] Update to v2019.01
Signed-off-by: Squareys <squareys@googlemail.com>
* [magnum] Update to v2019.01, remove shapes and glutapplication features
These features were deprecated in earlier versions and now removed.
Signed-off-by: Squareys <squareys@googlemail.com>
* [magnum-(extras,plugins,integration)] Update to v2019.01, add imgui feat.
Signed-off-by: Squareys <squareys@googlemail.com>
* [magnum] Improve dependencies to corrade
Signed-off-by: Squareys <squareys@googlemail.com>
* [magnum-plugins] Add faad2audioimporter
- faad2 is currently not in vcpkg, hence commented out
- The commented out empty line is required to have vcpkg parse the file
without issues
Signed-off-by: Squareys <squareys@googlemail.com>
* [ampl-mp] Add new port
This port aims for solver development, so it does not build any
existing solver modules.
The shared library build is disabled for now because it has some
problem on Windows.
closes: #4491
* [ampl-mp] Disable MATLAB MEX support
* fix library exports for iconv on macOS, add Threads dependency for cairo, add the iconv, intl and Threads dependencies to harfbuzz
* updated portfile from harfbuzz fix
* fix indenting and order
* fix portfile
* [gettext] Add unofficial-gettext targets
* [gettext] Fixup for Linux
This is so that you can derive a class from things like `rocksdb::AssociativeMergeOperator`. Perhaps a more "correct" patch would be to see if rtti is enabled for other builds, and enable it here, though it's not clear to me if that's feasible, so I'm suggesting this.
* add support for building hyperscan
* patch hyperscan to work with latest boost
* correct hyperscan source dir name
* speed the build by depending on only the parts of boost we actually need
* make python3 support release-only builds
* upgrade to hyperscan 5.1.0
* remove redundant disambiguate patch
* Add comments about how to build
* restore original x64-windows-static.cmake
* [hyperscan] Check linkage
* [python3] Allow DLLs without LIBs
* [hyperscan] Only static library
* [python3] Revert changes that cause regression in boost-python
* [python3] Revert changes that cause regression in boost-python
* Added cmake targets for nlohmann-json
* [nlohmann-json] Use vcpkg_from_github()
* [nlohmann-json] Enable find_package(nlohmann_json)
* Removed unneeded renaming
* [nlohmann-json] Revert use of vcpkg_from_github() due to archive size of >100MB
* [nlohmann-json] Fix missing natvis
Netcdf and HDF are important _formats_ in the scientific domains (lots of satellite data come in nc or hdf) so having them by default in GDAL seems the most natural.
* [systemc] Add new port (Microsoft#5250)
* [systemc] Fix debug targets
* [systemc] Use vcpkg_extract_source_archive_ex()
* [systemc] Disable vcpkg_test_cmake() for now due to issues
* [systemc] Fixup cmake targets
* Remove the libmysql dependency
The fact that it's not possible to build libmysql for x86 is a source of troubles for projects that have GDAL as a dependency. Though it's possible to build GDAL separately for x64 & x86 this is very confusing and breaks other builds. For example I'm often crossing this issue when trying to build a local GMT package. So the best is to just remove it as a dependency.
```
C:\programs\compa_libs\vcpkg>vcpkg install gmt:x86-windows --recurse
The following packages will be built and installed:
* gdal[core,mysql-libmysql]:x86-windows
gmt[core]:x86-windows
* libmysql[core]:x86-windows
Additional packages (*) will be modified to complete this operation.
Starting package 1/3: libmysql:x86-windows
Building package libmysql[core]:x86-windows...
CMake Error at C:/programs/compa_libs/vcpkg/ports/libmysql/portfile.cmake:10 (message):
Oracle has dropped support in libmysql for 32-bit Windows.
```
* Revert "Remove the libmysql dependency"
* Remove the libmysql dependency
The fact that it's not possible to build libmysql for x86 is a source of troubles for projects that have GDAL as a dependency. Though it's possible to build GDAL separately for x64 & x86 this is very confusing and breaks other builds. For example I'm often crossing this issue when trying to build a local GMT package. So the best is to just remove it as a dependency.
```
C:\programs\compa_libs\vcpkg>vcpkg install gmt:x86-windows --recurse
The following packages will be built and installed:
* gdal[core,mysql-libmysql]:x86-windows
gmt[core]:x86-windows
* libmysql[core]:x86-windows
Additional packages (*) will be modified to complete this operation.
Starting package 1/3: libmysql:x86-windows
Building package libmysql[core]:x86-windows...
CMake Error at C:/programs/compa_libs/vcpkg/ports/libmysql/portfile.cmake:10 (message):
Oracle has dropped support in libmysql for 32-bit Windows.
```
* Update boost to 1.69. Fixes#5023
- Remove boost-signal (rationale: deprecated and removed in 1.69)
- Remove boost-hana workaround by xiangfan-ms #4186#3579 (rationale: 1.69 incorporates required fixes in hana master branch; builds fine in VS2017 15.9)
* [boost-signals] Restore signals from boost 1.68 because it is required by wt
* [boost-vcpkg-helpers] Improve dependency detection
* [boost-iostreams] Revert to cmake build to enable dependencies
* [boost-iostreams] Fix use of zstd in static builds
* [libtorrent] Update to 1.2.0
* [wt] Update to 4.0.5
* [boost-safe-numerics] [boost-vcpkg-helpers] Added safe_numerics; slight mod to regex match
* [hpx] Backport boost 1.69 fixes
* [libtorrent] Clean up cmake file installation
* [boost-type-traits] Fix MSVC arm
* libass dynamic library
it relies on fribidi, harfbuzz, fontconfig, libiconv
* Remove references to iconv and fontconfig
Because this port is currently Windows-only (Fribidi doesn't build on
Linux with vcpkg because of an issue with Meson), we don't want to worry
about FontConfig; DirectWrite should be enough.
* Freetype is important!
* Remove unnecessary fontconfig -- for real
* Remove unnecessary comment
* [BZip2] Fix Unix Dynamic Builds
* [bzip2] Bump control version
* [libgta] Only build one flavor
* [vcpkg_configure_cmake] Resolve#2375 since we now use CMake >= 3.11
* [json-spirit] Add initial port
* Modify PR to using pn85, enable some features
json-spirit is a weird case because the main repo that package managers
tend to use appears to not be actively maintained. Nevertheless, vcpkg
tries to use the "official" source of a repository if at all possible.
In order to do this, I've made the following changes to the PR:
- I've changed the upstream REPO, REF, and SHA512.
- I've passed options to disable building the demos and tests
- I've made the project build only as a static library.
The reason for this is that vcpkg does not support dynamical linkin.
Another reason is that the library as it is in the repo does not have
all the definitions correctly annotated.smlee-hdactech has made a PR
against the major upstream repo to fix this, butwe are going to defer
to the main repo until that change is accepted.
* patch for building dll.
* [xsimd] Add new port
* Add a header include guard to xsimd's algorithm.hpp
algorithm.hpp is not guarded by a header include guard, which can cause
difficulty for users. xhihaoy asked that we be sure to include this
patch because it caused a lot of pain.
* [amqpcpp] Add new port
* Force static linkage for amqpcpp
The documentation for this library says that shared linkage on Windows
is unsupported. Furthermore, the build currently errors out on Windows
because shared DLLs are not placed correctly.
vcpkg only supports static linkage on Linux and macOS, so I'm
categorically making amqpcpp build with static linkage on all platforms.
This unifies the platforms and gives consistent behavior
* add CONTROL and portfile for spectralib 0.7.0
* Rename spectralib -> spectra
I'm making this change based on the naming in the other repositories
that the library is packaged in (DPorts, FreeBSD Ports) and the
terminology used on the website. Code examples use Spectra* for the
prefixes, so I think using "Spectra" is better than "spectralib".
* [json-spirit] Add initial port
* Modify PR to using pn85, enable some features
json-spirit is a weird case because the main repo that package managers
tend to use appears to not be actively maintained. Nevertheless, vcpkg
tries to use the "official" source of a repository if at all possible.
In order to do this, I've made the following changes to the PR:
- I've changed the upstream REPO, REF, and SHA512.
- I've passed options to disable building the demos and tests
- I've made the project build only as a static library.
The reason for this is that vcpkg does not support dynamical linkin.
Another reason is that the library as it is in the repo does not have
all the definitions correctly annotated.smlee-hdactech has made a PR
against the major upstream repo to fix this, butwe are going to defer
to the main repo until that change is accepted.
* [selene] Update to v0.3.
Add libtiff as a dependency.
Update release hash.
Remove the `vcpkg_check_linkage` call, since the relevant symbols for
shared libraries are now explicitly exported.
* [selene] Update to v0.3.1.
Also reinstate linkage check.
* [aws-c-common] Add new port
* [aws-checksums] Add new port
* [aws-c-event-stream] Add new port
* [aws-sdk-cpp] Update port to version 1.7.38
* [aws-sdk-cpp] Add dependencies on aws-c-event-stream, aws-checksums
This port currently only works on Windows because the dependencies of
the port (fribidi in particular) don't build on non-Windows platforms.
I made patch modifications to the source code because the Visual C++ C
compiler does not allow for non-constant-expression array indicesto be
used to declare a static array.
I've verified that this (1) builds on Windows and (2) works with a
sample test that's included in the libass github repo.
The GeoTIFF library ships with its own FindPROJ4.cmake module. This file
interferes with the cmake files that are installed by vcpkg for the PROJ4
package. As a result, the debug version of libgeotiff gets incorrectly
linked to the release version of PROJ4. By removing FindPROJ4.cmake from
the source directory, the vcpkg's .cmake files are used instead,
allowing CMake to find the debug version of PROJ4 when building the
debug version of libgeotiff.
* added patch to freetype for assigning a value to macros like HAVE_UNISTD_H as using newer macOS Frameworks fail to compile if they have no value
* [freetype] Modernize. Bump control version
* Add public-preview feature to azure-iot-sdk-c suite of packages
* [azure-c-shared-utility][azure-iot-sdk-c] Remove unneeded repeat dependencies in feature
* update qt5-base to 5.12
* update modules to 5.12 (qt5-declarative and dependents don't work yet)
* fix qt5-declarative for 5.12 and some hashes.
* add qt5-declarative as dependency for qt5-graphicaleffects
* Arrow update
* Add version info
* Forgot to apply patch
* Whitespace
* Fix patch
* Fix patch
* Version which works on Linux
* Fix problem finding ZSTD on x64-windows
* Fix zlib dependency finding
* Fix renaming on Windows
* Include missing boost-format dependency
* Linux debug build picks up debug libs now
* Make windows work again
* Remove boost-format dependency
* Error if dependencies not found rather than attempt to download them
* [parquet] Turn parquest into a stub
* [sciter] Update to 4.2.0.6064
* [sciter] Update to 4.2.1.6085
* [sciter] Update to 4.2.2.6090
* [sciter] Update to 4.2.3.6108
* [sciter] Update to 4.2.4.6184
* Update sciter to 4.2.6.9
* use d suffix for debug libs, improve compatibility with official FindJPEG.cmake (cmake 3.13)
* [libjpeg] fix also non-static case
* [libjpeg-turbo] bump version
* [libjpeg-turbo] fixes for regression on Linux/macOS
* [libjpeg-turbo] fix regression in qt5-base
* [linkjpeg-turbo] force CI test
* fix fmilib not building on x64-windows
* [fmilib] bump dash version in control file
This will force a rebuild when updating
* only search for MSVC generator on windows
On newer CMake's, the following errors are reported:
```
CMake Warning (dev) at .... /vcpkg/installed/x64-osx/share/c-ares/c-ares-config.cmake:32 (if):
if given arguments:
"ON"
An argument named "ON" appears in a conditional statement. Policy CMP0012
is not set: if() recognizes numbers and boolean constants. Run "cmake
--help-policy CMP0012" for policy details. Use the cmake_policy command to
set the policy and suppress this warning.
```
Which corresponds to the following configured code:
```
if(ON)
set(c-ares_STATIC_LIBRARY c-ares::cares_static)
endif()
```
* Use default library type for the platform instead of specifying SHARED
* [graphqlparser] Bump control version and slight cleanup
* [vcpkg_find_acquire_program] Correct for previous corruption of CI
Using an external fmt lib should cause the spdlog::spdlog target to have
a dependency on fmt lib - so that a consuming project does not need
to call find_package(fmt) and target_link_libraries(... fmt::fmt).
To this end a new cmake option SPDLOG_FMT_EXTERNAL is introduced which
makes spdlog depend on fmt lib and defines the SPDLOG_FMT_EXTERNAL macro
to avoid using the bundled fmt lib. The value of SPDLOG_FMT_EXTERNAL is
also stored in the installed spdlogConfig.cmake and if it is ON
find_dependency() is used to ensure the fmt::fmt target is imported.
gcc 7 deprecates the use of varargs.h in favour of stdarg.h.
mpfr already fixes it, but the fix is enabled if stdarg is available.
This patch adds the test on CMakeLists.txt and enables the fix if
needed.
* xalan-c: Remove use of obsolete unary_ and binary_function for C++17 compatibility
* [xerces-c] Improve port stability by adding control over ICU dependency
* move cmake files to proper location, and modify path in contents accordingly
* workaround cmake script to find openjpeg package, and include openjpeg in Build-Depends list
* [gdcm2] Use vcpkg_fixup_cmake_targets()
* fix pango for macOS: added an already parsed config.h.unix and moved win32 code conditionally for platform
* use a list for better viewing wich source files are added
* only define HAVE_CORE_TEXT on macOS
* [pango] Modernize. Bump control version. Fix configure into source directory.
* [cppgraphqlgen] Add a port for Microsoft/cppgraphqlgen
* Update hash for replacement tag
* Change the github ref to a commit ID to invalidate cached downloads
* Upgrade to cppgraphqlgen v0.6
* CppZMQ library version leveraged to 4.3.0.
* [cppzmq] Disable building tests. Remove vendored FindZeroMQ since our copy already supports the targets.
* [curl] Split the OpenSSL and SChannel backends into separate features
* [curl] Add mbedTLS backend as a feature
* [curl] Make winssl the default for Windows. Remove need for compatibility workaround.
* [curl] Continue to use openssl in UWP because PSecurityFunctionTableA is not available.
* [wxwidgets] move setup.h to include
Also remove lib/mswu and debug/lib/mswud. They only contain setup.h.
Fixes#3180.
Closes#4251.
* [wxwidgets] remove include/msvc directory
The only thing in include/msvc is include/msvc/wx/setup.h which is a
"wrapper" around the real setup.h. The wrapper setup.h is hard-coded to
include the real setup.h from lib but since the real setup.h is now in
include/wx the wrapper has become useless.
* [wxwidgets] Fix osx build. Slight modernization.
* [xmsh] add support for xmsh library
- xmsh is a mesh document format. xmsh library is the reference
implementation for the format.
* [xmsh] removed extraneous file command
* [xmsh] update version to v0.2.3
* [xmsh] upgrade to version v0.3.1
* [xmsh] update to version v0.4
* + fix python executable packaging
* [vcpkg_find_acquire_program] Fix PYTHON3 on non-Windows
* [Vulkan] Add a vulkan port based on the cuda port
* Add VULKAN_SDK env variable to whitelist
* * Added some additional diagnostic information
* Corrected if NOT exists statement
* New package: cutelyst2
A C++ Web Framework built on top of Qt, using the simple approach of Catalyst (Perl) framework.
* [cutelyst2] Rearrange files; prefer ninja.
* Update google-cloud-cpp to 0.3.0.
A new release of google-cloud-cpp, with some improvements in packaging
that simplified the port files.
* [google-cloud-cpp] Fix OSX builds
* [bzip2] Fix tools install directory
Change to install tools in port name directory. (vcpkg/installed/<triplet>/tools/bzip2)
* [bzip2] Modernize and bump control version
* [portaudio] Enable debug output.
This is enabled by default in the original MSVC project in the PortAudio
distribution, but not in their CMake build. This commit fixes the
regression.
I don't think there are any performance concerns with this, as PortAudio
(AFAICT) does not log from performance-critical code.
Note that this change might still be somewhat controversial, because
PortAudio will by default dump its debug output directly to stderr,
which is not particularly nice. However, the alternative is not having
any way for end users to troubleshoot PortAudio issues at all, which
IMHO is worse. Applications can always call
`PaUtil_SetDebugPrintFunction()` to redirect the output away from
stderr, and I encourage them to do so.
* [portaudio] Bump control version
* + added support for dascandy/pixel
* [pixel] Cleanup comments
* [SDL2] Add SDL2::SDL2{,-static} wrapper
* [pixel] update to v0.2
- fixes linking problem against SDL2
* [pixel] Bump control version
* [portaudio] Add pa_win_waveformat.h to public includes.
pa_win_wasapi.h includes pa_win_waveformat.h, so it has to be
distributed for pa_win_wasapi.h to be usable.
* [portaudio] Bump control version
* Add initial support for RapidCheck
Create a portfile for the C++ property-based testing library RapidCheck.
This port is relatively straight forward but there are a couple of caveats:
* The current port uses a branch from a fork that I modified as the CMake installation logic had some problems.
The library was almost exclusively used by being added as a subfolder so the CMake export machinery was probably not well tested.
* The changes are going to get merged into the main repository eventually and then it would be possible to change where the source
code is downloaded.
I am unaware when the main RapidCheck library maintainer will accept my pull request, so it is better to have this library in VCPKG
sooner, rather than later.
* Update the RapidCheck portfile to download the library sources for the main repository
* Since the maintainer of RapidCheck very quickly merged the required CMake modifications
to enable the proper installation of the library, it is better to now use the official
repository's master branch.
* Change the versioning of the library to be based on the date of the git reference chosen
since rapid check does not have an explicit versioning scheme.
* Update rapidcheck and modify configuration step
* [rapidcheck] Tidy
* fdlibm init
* fix links
* Fix different hashes creation with google host
* Move functions to script
* Fix documentation
* [vcpkg_from_git] Add SHA512 argument, switch to zip to better support Windows.
* [fdlibm] Trigger rebuild
* [vcpkg_from_git] Use FETCH_HEAD reference to support tags
* [rocksdb] Update to v5.15.10
* [rocksdb] Cleanup
* [rocksdb] Fix cmake targets name.
* [toolchains] Set NDEBUG in release
* [rocksdb] Disable vcpkg_test_cmake due to misbehavior in x64
* [tiff] Use lzma on linux
* [libgeotiff] Install doc files in the correct place
The original portfile put documentation (authors, readme, etc.) inside
of vcpkg/installed/<triplet>/doc. This commit modifies the portfile such
that the files are installed to
vcpkg/installed/<triplet>/share/libgeotiff/doc instead.
* [libgeotiff] Bump version
* [libgeotiff] Avoid file(INSTALL)
* Add SKIP_CLEAN option to vcpkg_install_msbuild
Also add vcpkg_clean_msbuild function to factor out clean
logic and allow re-use in portfiles.
* xalan-c: Correct header globbing
* vcpkg_install_msbuild: SKIP_CLEAN documentation correction
* [xalan-c] Add explicit check for localmsgindex header.
[docs] Regenerate
* [sdl1][sdl1-net] Add new port for SDL v1.2.15 & SDL_net v1.2.8
* Version 1.2 is still required for building DOSbox
* [sdl1][sdl1-net] Revise to install to expected paths
* Headers go in include\ instead of include\SDL1
* Use updated .sln and .vcxproj files
* [sdl1] Modernize
* Include libpq-events header file
Missing the libpq events header file. This file is needed to use the asynchronous libpq interfaces.
* [libpq] Bump control version
* [xerces-c] Update to 3.2.2
* [xalan-c] Build against xerces-c 3.2.2 and include extra patches
* [xalan-c] Use new vcpkg_extract_source_archive_ex()
* [xerces-c] Use vcpkg_fixup_cmake_targets
* [xerces-c] Disable vcpkg_test_cmake due to misbehavior in x64
* [xerces-c] Remove bin directory when building static
* [xalan-c] Modernize
* Correctly configure GTK3 to reflect recomended release settings
GTK/Glib packaging guidelines recomends a release build to use debugging
facilities set to level 'minimum'. Disabling this is officially not
recomended for stable releases. For more info read the following link,
the section `--enable-debug`.
https://developer.gnome.org/glib/stable/glib-building.html#extra-configuration-options
* [gtk] Bump control version
* [libusb] Update to 1.0.22
* Update version to v1.0.22 tag instead of non-release commit hash.
* Remove libusb-usbdk output library.
The two Windows backends have been unified. The UsbDk backend is
selectable at runtime using the function libusb_set_option().
* Remove libfreenect2 patch (changes integrated upstream).
* Tidy up portfile.
* [libusb] Use vcpkg_install_msbuild()
* [libfreenect2] Add missing libjpeg-turbo dependency
* making sure there is a leading <compileflags> even if CMAKE_CXX_FLAGS do not start with a whitespace
* making sure there is a leading <compileflags> even if CMAKE_CXX_FLAGS do not start with a whitespace
* full support for CXXFLAGS, CFLAGS and LDFLAGS throughout the boost-modular-build-helper
* [boost-modular-build-helper] Bump version and use CMAKE_* macros since the VCPKG_* macros are not passed in
* [boost-type-erasure] Depend on boost-thread even in UWP
* [orc] add apache orc library
* [orc] Simplify patch. Note in usage that no targets are available. Don't build tools on windows because they are not supported.
* [orc] Move tools to tools/orc/ to fix Linux build.
* [orc] fix tools ouput folder
* Fix crc32c CMake configuration directory.
The port was installing the CMake configuration files for Crc32c in
share/crc32c/Crc32c/Crc32cConfig.cmake, where find_package() cannot
find them. They should be installed in share/crc32c/Crc32cConfig.cmake.
* [crc32] Lowercase share directory
