So, what are the things needed to make a new Boost available to Jenkins and Teuthology? If I have access to do them I'll happily do them. (Or if there are time consuming parts that don't require special access I can do those and hand over the pieces.)

there are some developer docs about boost in https://docs.ceph.com/en/latest/dev/continuous-integration/#build-dependencies

Hi Adam, as Casey pointed out. in addition to precompiling boost, we also need to upload the tarball to https://download.ceph.com/qa/.

Jenkins, retest this please.

@djgalloway Could you please build/upload debs for Boost? It think that's the only thing blocking this PR (and I'm pretty sure I don't have access) and this PR is the only thing blocking C++20.

jenkins retest this please

@adamemerson I built packages. This was my first time attempting it so we may still hit some snags.

https://shaman.ceph.com/repos/libboost/master/892ab89e76b91b505ffbf083f6fb7f2a666d4132/default/256966/

@adamemerson I built packages. This was my first time attempting it so we may still hit some snags.

https://shaman.ceph.com/repos/libboost/master/892ab89e76b91b505ffbf083f6fb7f2a666d4132/default/256966/

Thank you very much!

892ab89e76b91b505ffbf083f6fb7f2a666d4132

Done! Thank you.

Looks like https://github.com/ceph/ceph/blob/main/win32_deps_build.sh#L27-L28 also need to be bumped

Looks like https://github.com/ceph/ceph/blob/main/win32_deps_build.sh#L27-L28 also need to be bumped

Done, thank you.

../build.deps/mingw/boost/include/boost/align/aligned_allocator.hpp: In member function 'void boost::alignment::aligned_allocator<T, Alignment>::deallocate(boost::alignment::aligned_allocator<T, Alignment>::pointer, boost::alignment::aligned_allocator<T, Alignment>::size_type)':
../src/include/compat.h:352:27: error: '_aligned_free' is not a member of 'boost::alignment'; did you mean 'aligned_free'?
  352 | #define aligned_free(ptr) _aligned_free(ptr)
      |                           ^~~~~~~~~~~~~

turns out aligned_free is a member of boost::alignment, so using a hackish #define to override the default behavior of aligned_free() fails the build.

i am proposing a fix to address this. see #47212

jenkins test windows

jenkins test make check

I'm okay with this as long as make check passes. Please don't merge until Friday morning at the earliest, however. I want to be around for fallout if necessary.

And thank you so much @tchaikov for your help. I absolutely would not have figured out that compilation issue you fixed in #47212. We miss you!

i am proposing #47217 to address this issue. the proposed PR includes the commits in this PR to verify my change.

@adamemerson could you update make-dist accordingly? otherwise we'd have FTBFS like

Done!

jenkins test windows

jenkins test make check

https://shaman.ceph.com/builds/ceph/pr-46554-kefu/0c4e48e672b8894c273bcd779235bcdcfd9f77f5/

@tchaikov The RGW suite is seeing valgrind issues that may be related to boost. Not due to 1.79 explicitly but we were wondering if you could take a look at https://tracker.ceph.com/issues/56500#note-2. Do you understand what's going on there?

@djgalloway will take a look tomorrow. just saw your comment right before i am about to call it a day.

essentially, the rgw/verify suite needs a way to run its valgrind jobs against builds (both ceph and boost) with the valgrind flag enabled

would it be possible to:

• build these ceph-libboost- packages with the valgrind=on flag
• set WITH_BOOST_VALGRIND=ON for all ceph builds that link to these packages

where would this logic need to go? i see install_boost_on_ubuntu() in https://github.com/ceph/ceph/blob/main/install-deps.sh#L169, but nothing for centos

cc @yuvalif

@cbodley i see you scheduled couple rgw:verify tests with pr-46554-kefu. they all completed without reporting issues. but according to the jenkins build log

CEPH_EXTRA_CMAKE_ARGS=-DALLOCATOR=tcmalloc
...
	cd obj-x86_64-linux-gnu && cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=None -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_INSTALL_LOCALSTATEDIR=/var -DCMAKE_EXPORT_NO_PACKAGE_REGISTRY=ON -DCMAKE_FIND_PACKAGE_NO_PACKAGE_REGISTRY=ON "-GUnix Makefiles" -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_AUTOGEN_VERBOSE=ON -DCMAKE_INSTALL_LIBDIR=lib/x86_64-linux-gnu -DWITH_JAEGER=ON -DWITH_SYSTEM_UTF8PROC=ON -DWITH_OCF=ON -DWITH_LTTNG=ON -DWITH_MGR_DASHBOARD_FRONTEND=OFF -DWITH_PYTHON3=3 -DWITH_CEPHFS_JAVA=ON -DWITH_CEPHFS_SHELL=ON -DWITH_SYSTEMD=ON -DCEPH_SYSTEMD_ENV_DIR=/etc/default -DWITH_GRAFANA=ON -DWITH_RBD_RWL=ON -DWITH_RBD_SSD_CACHE=ON -DCMAKE_INSTALL_LIBDIR=/usr/lib -DCMAKE_INSTALL_LIBEXECDIR=/usr/libexec -DCMAKE_INSTALL_SYSCONFDIR=/etc -DBOOST_J=48 -DALLOCATOR=tcmalloc ..
...
-- BUILDING Boost Libraries at j 48
-- boost (1.79.0 >= 1.79) already in /build/ceph-17.0.0-13785-g0c4e48e6/src/boost
...
cd /build/ceph-17.0.0-13785-g0c4e48e6/src/boost && CC=/usr/bin/cc CXX=/usr/bin/c++ ./b2 -j48 -d0 --user-config=/build/ceph-17.0.0-13785-g0c4e48e6/obj-x86_64-linux-gnu/user-config.jam toolset=gcc python=3.8 headers stage variant=release threading=multi link=static address-model=64 "cxxflags=-fPIC -w"

the packages were not built with valgrind=on or WITH_BOOST_VALGRIND=ON. so i am afraid that your analysis cannot explain the completed runs on ubuntu?

the ceph-libbooost- packages are not involved in qa runs. they are only used when performing "make check" tests performed by jenkins.

if WITH_BOOST_VALGRIND=ON is proved to be the cure. we need add it at https://github.com/ceph/ceph-build/blob/4aa26880bb7917fa65b5ed11242b4f59aae91a77/scripts/build_utils.sh#L831

cc @yuvalif

i tried with the change. did not help. e.g.
http://pulpito.front.sepia.ceph.com/yuvalif-2022-07-28_04:22:02-rgw:verify-wip-yuval-fix-lua-post-ctx-distro-default-smithi/
http://pulpito.front.sepia.ceph.com/yuvalif-2022-07-28_04:20:01-rgw:verify-wip-yuval-lua-filter-distro-default-smithi/

so i am afraid that your analysis cannot explain the completed runs on ubuntu?

thanks @tchaikov. i didn't expect WITH_BOOST_VALGRIND to be on there, i just wanted to test against ubuntu with recent builds, and yours could help rule out the boost version also. that first passing run didn't include any valgrind jobs at all, so i had to reschedule in https://pulpito.ceph.com/cbodley-2022-07-26_14:01:08-rgw:verify-pr-46554-kefu-distro-default-smithi/. that result confirmed that the valgrind issues are effecting ubuntu also - i reported on this in https://tracker.ceph.com/issues/56500#note-8

that first passing run didn't include any valgrind jobs at all

ahh, i see. so we need to go with WITH_BOOST_VALGRIND=ON then.

@tchaikov ISTR you asking to build boost for arm64 too. Done. https://shaman.ceph.com/repos/libboost/master/892ab89e76b91b505ffbf083f6fb7f2a666d4132/default/257913/

@tchaikov ISTR you asking to build boost for arm64 too. Done. https://shaman.ceph.com/repos/libboost/master/892ab89e76b91b505ffbf083f6fb7f2a666d4132/default/257913/

@djgalloway yes, thank you! to be on par with amd64 nodes. Thank you!