Merge tag darwin/darwin-008 on darwin/trunk. #252
Merged
Conversation
This file contains hidden or 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
Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
The static version (`libdispatch.a`) that can be compiled using
`./configure --enable static=yes` was missing all the Swift overlay symbols
(everything in `swift_overlay.o`). The reason for that is that the linker is
invoked through libtool which wants `.lo` files but the Swift overlay got passed
as a `.o` file. This first of all leads to this warning:
*** Warning: Linking the shared library libdispatch.la against the non-libtool
*** objects [...]/swift_overlay.o is not portable!
And the result is that libtool doesn't include `swift_overlay.o` when putting
together the static library `libdispatch.a`.
This patch fixes this problem is a pretty crude way. The real problem is that
libtool doesn't understand Swift. So it can't be used when compiling
`swift_overlay.o`. In order to still get a `.lo` file this patch tricks libtool
into generating one from the `swift_overlay.o` generated by `swiftc`. It's very
hacky but I'm not sure what else to do.
Signed-off-by: Daniel A. Steffen <[email protected]>
fix libdispatch.a to include Swift overlay symbols Signed-off-by: Daniel A. Steffen <[email protected]>
Collection of small fixes to dispatch-806 merge to allow code to compile/link/run on Linux. Signed-off-by: Daniel A. Steffen <[email protected]>
…rge-master-linux-fixes Linux fixes for dispatch-806 merge Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
provide fallback definitions for all API_* macros Signed-off-by: Daniel A. Steffen <[email protected]>
add dependency on linux-libc-dev to INSTALL.md for other architectures Signed-off-by: Daniel A. Steffen <[email protected]>
remove unused dependency on linux/membarrier.h for intel to fix CI build Signed-off-by: Daniel A. Steffen <[email protected]>
Also reorder all vtable entries to be in the same order (push, invoke, wakeup). Reenable dispatch_context_for_key. SR: SR-4039 Signed-off-by: Daniel A. Steffen <[email protected]>
queue_specific_queue is missing its do_push vtable entry Signed-off-by: Daniel A. Steffen <[email protected]>
Collection of small fixes to eliminate compiler warnings when building libdispatch and its tests. Signed-off-by: Daniel A. Steffen <[email protected]>
fixes for compiler warnings Signed-off-by: Daniel A. Steffen <[email protected]>
(Radar rdar://problem/30518517) Signed-off-by: Daniel A. Steffen <[email protected]>
(Radar 30699743) Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
…ting the fact that a relative or non-existent path is invalid. (Radar 31115994) Signed-off-by: Daniel A. Steffen <[email protected]>
… model and does not require membarrier. Signed-off-by: Daniel A. Steffen <[email protected]>
Fix compile error on s390x in shims/lock.h Signed-off-by: Daniel A. Steffen <[email protected]>
libdispatch has an umbrella header, dispatch.h, that should be used for including all of the other headers. This is enforced via __DISPATCH_INDIRECT__. Since it isn't legal to include one of the other headers on their own, it isn't logical to have submodules for them. Moreover, submodules *should* have local visibility (be unaware of the context they're included from); meaning that __DISPATCH_INDIRECT__ will never be defined. Logically, all of the headers are part of a single module/interface. Patch by me & Duncan P. Exon Smith Signed-off-by: Daniel A. Steffen <[email protected]>
libdispatch: Avoid submodules for headers that should not be included Signed-off-by: Daniel A. Steffen <[email protected]>
(Radar 31585625) Signed-off-by: Daniel A. Steffen <[email protected]>
This is far from complete, but is sufficient to build a Linux version of libdispatch. It shows what a potential cmake based build system could look like, and if desired can be completed to build all the various flavours with cmake. Signed-off-by: Daniel A. Steffen <[email protected]>
build: add a cmake based build system Signed-off-by: Daniel A. Steffen <[email protected]>
Building swift-corelibs-libdispatch on powerpc64le under Linux lead to the following build failure: /bin/bash ../libtool --tag=CXX --mode=compile /home/ubuntu/swift-dev/build/buildbot_incremental/llvm-linux-powerpc64le/bin/clang++ -DHAVE_CONFIG_H -I. -I../config -I.. -I.. -I../private -DDISPATCH_USE_DTRACE=0 -I../libpwq/include -Wall -fvisibility=hidden -momit-leaf-frame-pointer -isystem /usr/include/bsd -DLIBBSD_OVERLAY -fblocks -I../src/BlocksRuntime -std=gnu++11 -fno-exceptions -O2 -c -o libdispatch_la-block.lo `test -f 'block.cpp' || echo './'`block.cpp libtool: compile: /home/ubuntu/swift-dev/build/buildbot_incremental/llvm-linux-powerpc64le/bin/clang++ -DHAVE_CONFIG_H -I. -I../config -I.. -I.. -I../private -DDISPATCH_USE_DTRACE=0 -I../libpwq/include -Wall -fvisibility=hidden -momit-leaf-frame-pointer -isystem /usr/include/bsd -DLIBBSD_OVERLAY -fblocks -I../src/BlocksRuntime -std=gnu++11 -fno-exce ptions -O2 -c block.cpp -fPIC -DPIC -o .libs/libdispatch_la-block.o In file included from block.cpp:32: In file included from ./internal.h:628: In file included from ./shims.h:171: ./shims/lock.h:550:3: error: use of undeclared identifier 'DISPATCH_INTERNAL_CRASH' DISPATCH_INTERNAL_CRASH(errno, "sys_membarrier not supported"); ^ 1 error generated. Makefile:701: recipe for target 'libdispatch_la-block.lo' failed make[2]: *** [libdispatch_la-block.lo] Error 1 make[2]: Leaving directory '/home/ubuntu/swift-dev/swift-corelibs-libdispatch/src' Makefile:541: recipe for target 'all' failed make[1]: *** [all] Error 2 make[1]: Leaving directory '/home/ubuntu/swift-dev/swift-corelibs-libdispatch/src' Makefile:457: recipe for target 'all-recursive' failed make: *** [all-recursive] Error 1 Include ordering in internal.h is tightly constrained, so open-code the macro to avoid the dependency problem. Signed-off-by: Andrew Jeffery <[email protected]> Signed-off-by: Daniel A. Steffen <[email protected]>
lock: Avoid use of undefined DISPATCH_INTERNAL_CRASH Signed-off-by: Daniel A. Steffen <[email protected]>
Introduce support for building the swift SDK overlay into the CMake based build system. This was the one piece which was missing from the original cmake port of the build system. Signed-off-by: Daniel A. Steffen <[email protected]>
build: attempt to add swift support to cmake Signed-off-by: Daniel A. Steffen <[email protected]>
The observed bug is that signals sent to the main thread after it called dispatch_main were not triggering the event handler of a DispatchSourceSignal that had been previously created and activated to handle that signal. This patch addresses the problem by: - export _dispatch_sigmask() as the way to block signals in a process using dispatch - call it when dispatch_main() is called - each time dispatch registers a signal handler, register a sigaction that will catch signals for threads with wrong masks and will fix the thread configuration then raise the signal again by directing a pthread_kill to the manager thread so that it is delivered to the signalfd as expected. Patch by Pierre Habouzit <[email protected]> from PR-231 with minor compilation fixes by Dave Grove. Signed-off-by: Daniel A. Steffen <[email protected]>
SR-4201: DispatchSourceSignal not working on Linux Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
This introduces a new macro, `DISPATCH_PTR_SIZE` which is defined much like `LLVM_PTR_SIZE`. We can gracefully fallback to alternate means of checking the width of the pointer and use this rather than `__LP64__` to determine if we are on a 64-bit host. Signed-off-by: Daniel A. Steffen <[email protected]>
Create a DISPATCH_PTR_SIZE macro and use it Signed-off-by: Daniel A. Steffen <[email protected]>
sys/cdefs.h and unistd.h are not always available. We already have checks for these, guard the inclusion appropriately. Signed-off-by: Daniel A. Steffen <[email protected]>
Rather than checking for non-Linux and including the target specific header, invert the check and include the linux target header if `__linux__` is defined. This makes it easier to target other platforms. Signed-off-by: Daniel A. Steffen <[email protected]>
Address some of the portability considerations for the build flags. This improves some of the conditions for building for Windows. Signed-off-by: Daniel A. Steffen <[email protected]>
Windows cleanups Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
This fixes a typo in the build system which is important because it fixes the dependency tracking and the output location for the swift doc that is generated for the swift SDK overlay. Signed-off-by: Daniel A. Steffen <[email protected]>
build: fix typo in build system Signed-off-by: Daniel A. Steffen <[email protected]>
Provide an implementation of the pthread_workqueue functionality (libpwq) as a fully integrated component of libdispatch. Integration of the workqueue implementation into the libdispatch code base simplifies the process of evolving the APIs between the two layers and thus prepares for future optimization and enhancements. The overall design is to augment libdispatch's existing pthread_root_queue option with periodic user-space monitoring of the status of a root queue's worker threads and to oversubscribe the workqueue when not enough workers are observed to be runnable when the queue has available work. Initially, the integrated pthread_workqueue is only enabled by default on Linux. The current monitoring implementation relies on Linux-specific code to dynamically estimate the number of runnable worker threads by reading /proc. Porting the monitoring code to non-Linux platforms would entail providing similar functionality for those platforms. Remove libpwq git submodule and autotools support for building libpwq from source as part of building libdispatch. Signed-off-by: Daniel A. Steffen <[email protected]>
implement pthread_workqueue within libdispatch Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
Signed-off-by: Daniel A. Steffen <[email protected]>
|
@swift-ci please test |
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
dgrove-oss
approved these changes
May 25, 2017
src/shims.h
Outdated
| @@ -41,6 +41,8 @@ | |||
| #if HAVE_PTHREAD_WORKQUEUES | |||
| #if __has_include(<pthread/workqueue_private.h>) | |||
| #include <pthread/workqueue_private.h> | |||
| #elif DISPATCH_USE_INTERNAL_WORKQUEUE | |||
| #include "event/workqueue_internal.h" | |||
There was a problem hiding this comment.
I think since you merged #248, the elif is dead code. But fine to merge as is to save time; I can double check after the merge and submit a PR to remove this if I'm right.
There was a problem hiding this comment.
duh, I had resolved the merge conflict the wrong way around, thanks for spotting
squashed the fix
Signed-off-by: Daniel A. Steffen <[email protected]>
das
force-pushed
the
das-darwin-008-merge-master
branch
from
7406bad to
81a67c9
Compare
|
@swift-ci please test |
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.
This includes the github PR's that we've pulled back into the Apple internal repository starting with the merge of PR #216 up to the merge of PR #209, along with any changes made to them during internal review and testing on Darwin:
#216 from dgrove-oss/das-darwin-libdispatch-806-merge-master-linux-fixes
#218 from apple/das-api-availability-build-fix
#221 from apple/das-disable-linux-membarrier-include
#223 from apple/eng/PR-SR-4039
#226 from dgrove-oss/fix-compiler-warnings
#227 from ktopley-apple
#230 from ktopley-apple
#232 from ktopley-apple
#234 from linux-on-ibm-z/fix_compile_error
#236 from bcardosolopes/master
#238 from ktopley-apple
#196 from compnerd/cmake
#233 from amboar/master
#243 from compnerd/swiftDispatch
#231 from dgrove-oss/SR-4201-signalsource
#244 from compnerd/ptrsize
#245 from compnerd/windows-cleanups
#246 from compnerd/build-dependencies
#206 from dgrove-oss/internal-pwq-impl
#209 from weissi/jw-fix-libdispatch.a