F40 Change Proposal: Boost 1.83 Upgrade (System-Wide)

Boost 1.83 upgrade

This is a proposed Change for Fedora Linux.
This document represents a proposed Change. As part of the Changes process, proposals are publicly announced in order to receive community feedback. This proposal will only be implemented if approved by the Fedora Engineering Steering Committee.


:link: Summary

This change brings Boost 1.83 to Fedora. This will mean Fedora ships with a recent upstream Boost release.

:link: Owner

:link: Detailed Description

The aim is to synchronize Fedora with the most recent Boost release. Because ABI stability is absent from Boost, this entails rebuilding of all dependent packages. This also entails the change owner assisting maintainers of client packages in decoding cryptic boost-ese seen in output from g++.

The equivalent changes for previous releases were Changes/F39Boost181, Changes/F37Boost178, Changes/F35Boost176, Changes/F34Boost175, Changes/F33Boost173, Fedora 30 Change, Fedora 29 Change, Fedora 28 Change, Fedora 27 Change, Fedora 26 Change, Fedora 25 Change, Fedora 24 Change, Fedora 23 Change and Fedora 22 Change.

:link: Benefit to Fedora

Fedora 40 includes Boost 1.83.

Fedora will stay relevant, as far as Boost clients are concerned.

Boost 1.83 and 1.82 do not bring any new components, but includes two new header-only libraries -

  • Compat: A repository of C++11 implementations of standard components added in later C++ standards, from Peter Dimov and contributors.
  • Mysql: A C++11 client for the MySQL database server, based on Boost.Asio, from Ruben Perez.

Boost 1.83 also includes many fixes and enhancements to existing components, including major and potentially breaking changes to

  • Multi-index Containers: Serialization now uses std::size_t instead of unsigned long to save and load the size of a multi_index_container (unsigned long is smaller than std::size_t in LLP64 data models). multi_index_container serialization class version has been bumped from 3 to 4 to reflect this change.
  • Timer: Disable deprecated headers unless BOOST_TIMER_ENABLE_DEPRECATED is defined.

Additionally the following breaking changes were introduced in Boost 1.82 and so also affect 1.83:

  • Atomic: Removed atomic::storage() accessors and atomic::storage_type types that were deprecated in Boost.Atomic 1.73. Users are recommended to use atomic::value() and atomic::value_type instead.
  • Geometry: The WKT output presentation of an empty polygon is now POLYGON() to make it consistent with other geometries
  • Math: Deprecated C++11 support.
  • Multiprecision: Deprecated C++11 support.
  • Process: Deprecated wait_for & wait_until in V1 - they will be removed in the future!

:link: Scope

  • Proposal owners:

    • Build will be done with Boost.Build v2 (which is the upstream-sanctioned way of building Boost)
    • Request a “f40-boost” build system tag (discussion):
    • Build boost into that tag (take a look at the build #606493 for inspiration)
    • Post a request for rebuilds to fedora-devel
    • Work on rebuilding dependent packages in the tag.
    • When most is done, re-tag all the packages to rawhide
    • Watch fedora-devel and assist in rebuilding broken Boost clients (by fixing the client, or Boost).
  • Other developers:

    • Those who depend on Boost DSOs will have to rebuild their packages. Feature owners will alleviate some of this work as indicated above, and will assist those whose packages fail to build in debugging them.
  • Release engineering: TODO (a check of an impact with Release Engineering is needed)

  • Policies and guidelines:

    • Apart from scope, this is business as usual, so no new policies, no new guidelines.
  • Trademark approval: N/A (not needed for this Change)

:link: Upgrade/compatibility impact

  • No manual configuration or data migration needed.
  • Some impact on other packages needing code changes to rebuild. Historically this hasn’t been too much of a problem and could always be resolved before deadline.

:link: How To Test

  • No special hardware is needed.
  • Integration testing simply consists of installing Boost packages (dnf install boost) on Fedora and checking that it does not break other packages (see below for a way to obtain a list of boost clients).

:link: User Experience

  • Expected to remain largely the same.
  • Developers building third-party software on Fedora may need to rebuild against the new Boost packages, and may need to adjust their code if the new Boost release is not source-compatible.

:link: Dependencies

Packages that must be rebuilt: $ dnf repoquery -s --releasever=rawhide --whatrequires libboost\* --disablerepo=* --enablerepo=fedora | sort -u

All clients: $ dnf repoquery --releasever=rawhide --archlist=src --whatrequires boost-devel --disablerepo='*' --enablerepo=fedora-source

:link: Contingency Plan

  • Contingency mechanism: Worst case scenario is to abandon the update and simply ship F40 with Boost 1.81, which is already in rawhide.

  • Blocks release? No

  • Blocks product? None

:link: Documentation

:link: Release Notes

1 Like

Is that on purpose that Boost is always lagging behind in Fedora?

Boost 1.84 was published last week (December 13th, 2023), and the proposed version for Fedora 40 is Boost 1.83.

I understand that the change of Boost version in Fedora is difficult, because depending package must be rebuild (and they may not be ready for that). But, according to the schedule for Fedora 40, should not this proposal be rebased on Boost 1.84?

See Boost - Fedora Project Wiki

This update is less than a year since the last update in rawhide, which is considerably less than the 18 months we aim for. Boost release frequently, and updating to every release is not practical given available resources.

We’ve already been doing test builds with Boost 1.83.0 and found problems that need to be patched in Fedora (e.g. I pointed out Fix cstdfloat numeric limits detection macro for libstdc++14 by mborland · Pull Request #1057 · boostorg/math · GitHub which isn’t fixed in 1.84.0 either). A new release will have new problems that need to be resolved, when we’ve already invested time in finding the problems in 1.83.0.

Well, I can see that the Debian project is struggling as well. See boost-defaults - Debian Package Tracker and Transition: boost-defaults.

If test builds with Boost 1.83.0 have already been started, I understand that it is too late to rebase on 1.84.

Let me know how I can help.

This change proposal has now been submitted to FESCo with ticket #3127 for voting.

To find out more, please visit our Changes Policy documentation.