Replacing util::function_nonser on std::function in hpx_init

[toktarev]

Proposed Changes

Use std::function in public APIs

Any background context you want to provide?

Use std::function in public APIs: #4987

[jenkins-cscs]

Can one of the admins verify this patch?

[hkaiser]

@toktarev thanks for looking into this! Much appreciated. What compatibility implications do you see for this patch? Should we consider introducing a compatibility layer?

[toktarev]

@hkaiser
Thank you very much for review.

I will add back-compatibility tests

[hkaiser]

BTW, thanks for adding the move semantics for the intermediate APIs.

[msimberg]

Thanks for doing this!

As for compatibility, I wonder if we could get by without that... Since we're currently using hpx::util::function_nonser (not the unique version), even if someone has explicitly created the entry point upfront and written out the type:

hpx::util::function_nonser<void()> = myentrypoint;
hpx::init(myentrypoint, ...);

even that will work. hpx::util::function_nonser can't be constructed from a noncopyable callable just like std::function so that stays the same.

[toktarev]

hpx::init: hpx::exception caught: Pools empty of resources are not allowed. Please re-run this application with allow-empty-pool-policy (not implemented yet): HPX(invalid_status)
../libs/full/resource_partitioner/tests/unit/named_pool_executor.cpp(164): test 'hpx::init(argc, argv, init_args) == 0' failed in function 'int main(int, char **)': '-1' != '0'
0 sanity checks and 1 test failed.
Base command is "/Users/runner/work/hpx/hpx/build/bin/named_pool_executor_test --hpx:threads=2"
Executing command: /Users/runner/work/hpx/hpx/build/bin/named_pool_executor_test --hpx:threads=2

        Start 299: tests.unit.modules.resource_partitioner.resource_partitioner_info
299/905 Test #299: tests.unit.modules.resource_partitioner.resource_partitioner_info ......................................................***Failed    0.31 sec
../libs/full/resource_partitioner/tests/unit/resource_partitioner_info.cpp(21): test 'std::size_t(4) == hpx::resource::get_num_threads()' failed in function 'int hpx_main()': '4' != '2'

I am not sure if it was caused by my changes

[hkaiser]

LGTM, thanks!

[toktarev]

macOS CI (Debug) / build (pull_request) failed

Was it caused by my changes or this is regular failure ?

[msimberg]

macOS CI (Debug) / build (pull_request) failed

Was it caused by my changes or this is regular failure ?

Regular failure, unfortunately. The macOS build is not particularly well taken care of, but nothing for you to worry about.

[msimberg]

retest

[toktarev]

#!/bin/bash -eo pipefail
cd /hpx/source/libs && shopt -s globstar # to activate the ** globbing
clang-format-9 --version
clang-format-9 -i **/*.{cpp,hpp}
git diff --exit-code > /tmp/modified_clang_format_files.txt
clang-format version 9.0.1-12 

Exited with code exit status 1
CircleCI received exit code 1

3 tests failed
  out of 597
tests.unit.modules.runtime_components.distributed.tcp.migrate_component
tests.unit.modules.runtime_components.distributed.tcp.migrate_polymorphic_component
tests.unit.modules.timed_execution.timed_parallel_executor

Is this something I should fix ?

[hkaiser]

#!/bin/bash -eo pipefail
cd /hpx/source/libs && shopt -s globstar # to activate the ** globbing
clang-format-9 --version
clang-format-9 -i **/*.{cpp,hpp}
git diff --exit-code > /tmp/modified_clang_format_files.txt
clang-format version 9.0.1-12 

Exited with code exit status 1
CircleCI received exit code 1

3 tests failed
  out of 597
tests.unit.modules.runtime_components.distributed.tcp.migrate_component
tests.unit.modules.runtime_components.distributed.tcp.migrate_polymorphic_component
tests.unit.modules.timed_execution.timed_parallel_executor

Is this something I should fix ?

These tests fail occasionally. The failures are unrelated to your changes.

[toktarev]

retest

[hkaiser]

This looks good as is. It can go in.

[msimberg]

retest