Improve compatibility with Dinkumware C++ headers #1008
Conversation
…CPPLIB version to avoid using things that aren't available in Dinkumware C++ headers.
|
@mhorowitz this is the folly side of facebook/react-native#21764 |
|
Looks like the build is failing for unrelated reasons (can't persist the ccache after the build finished successfully). |
|
To give some perspective, overall, we are not planning to support older versions of compilers. This includes clang 3.9 and msvc 2015. As one particular, clang 3.9 has no support for C++17, so that compiler will definitely not work once we decide to jump to C++17. We also want to avoid all the extra complexity, preprocessor conditionals, restrictions on coding patterns, etc., that arise from supporting older compilers. |
matthargett
changed the title
|
@yfeldblum sorry, my title was misleading. This is with clang 6 and 7, but using Dinkumware headers. I was able to bump the Dinkumware headers and eliminate a bunch of the __has_include noise. Take another look and let me know what you think. |
|
Note that when PRs are imported into the internal code review tool or landed into the repo, the PR commits are squashed into a single commit. |
| @@ -23,6 +23,10 @@ | |||
| #include <folly/portability/Unistd.h> | |||
| #include <folly/portability/Windows.h> | |||
|
|
|||
| #if __has_include(<pthread_np.h>) | |||
| #include <pthread_np.h> | |||
There was a problem hiding this comment.
@Orvid Do we want this include here or in the portability header?
…e they were used. Added explicit glog/logging.h include to the top of the relevant files, to match the style in folly/detail/IPAddressSource.h
|
Looks like the build failure is an unrelated regression caused by a different commit? |
There was a problem hiding this comment.
@yfeldblum has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
Note that I will internally reorder the glog includes to match our style. |
|
any issues landing it in phabricator? let me know if there's anything else I can do to get it merged. |
| // are found in stdlib.h. | ||
| #if __has_include(<malloc.h>) | ||
| // are found in stdlib.h and it errors when including malloc.h. | ||
| #if __has_include(<malloc.h>) && (!defined(__STDC__) || !__STDC__) |
There was a problem hiding this comment.
This (or the change to the portability malloc below) is causing malloc.h to no longer get included in our internal builds :(
If these two changes are dropped, the rest should be safe to land :)
Summary:
These are a set of minor changes to allow building and linking on older clang and toolchains that use Dinkumware headers. I adjusted and tested only the subset of folly used by React Native's CxxReact, Fabric, and TurboModules C++ sources.
Test plan:
I ran the build on my Ubuntu 18.x using both clang 3.9, clang 6.x, and GCC 8.x. I couldn't find a way to build and run the unit tests in cmake, so I spot checked several unit tests with a manual CLI build:
g++-8 -I ../vcpkg/installed/x64-linux/include/ -L../vcpkg/installed/x64-linux/lib -I. -I_build folly/test/DynamicConverterTest.cpp folly/test/common/TestMain.cpp -lgtest -lpthread -lglog -lgflags -ldouble-conversion -lssl -lboost_filesystem -lboost_program_options -lboost_regex -lboost_system -ldl -lstdc++ -liberty _build/libfolly.a ../vcpkg/installed/x64-linux/lib/libdouble-conversion.a