build: set minimum required Boost to 1.64.0 #22320

1. 
   fanquake commented at 7:48 AM on June 23, 2021: member

   Setting a newer minimum required Boost means we can remove the awkward header / compile check for Boost Process.

   If we don't do this, we should at-least make Boost Process being missing no longer a failure, otherwise anyone building using Boost < 1.64.0 would have to pass --disable-external-signer as well.

   The only system I can see that is affected here, (doesn't have new enough system packages) is Debian Oldstable. However, anyone compiling there, can use depends. They can also no-longer use the system GCC (6.0), and I'd assume would be using Clang 7, which would be the newest compiler available to them. It's extended, LTS support also end in 1 year from now, so anyone still using it should be considering upgrading.

   Debian Buster (Stable) has 1.67+, Ubuntu Bionic has 1.65+, any of the BSDs, recent Fedora, macOS etc all also have well and truly new enough Boost versions available.

   I think this is something we should just do for 22.0. If not, definitely for 23.0.

   Fixes #22319. Compiling Bitcoin Core should work, as windows.h will be included. Alternative to #22294. Would also close #22269. #19128 could be re-opened.

2. fanquake added the label Build system on Jun 23, 2021
3. build: set minimum required Boost to 1.64.0 2bf211696b
4. build: remove workaround for Boost and std::atomic df2c933217
5. build: remove check for Boost Process header

   Now that we require Boost 1.64.0+, Boost Process will be available.

   957f358427
6. fanquake force-pushed on Jun 23, 2021
7. 
   MarcoFalke commented at 8:07 AM on June 23, 2021: member

   Concept ACK

8. 
   hebasto commented at 10:32 AM on June 23, 2021: member

   Concept ACK.

9. 
   hebasto commented at 11:09 AM on June 23, 2021: member

   Boost Process was introduced in 1.64. Debian stretch has boost 1.62, other main distros have 1.65+, right?

   If so, will we have any benefit of bumping boost version to 1.65 (instead of 1.64)?

10. 
    fanquake commented at 11:33 AM on June 23, 2021: member

    If so, will we have any benefit of bumping boost version to 1.65 (instead of 1.64)?

    The only benefit might be being able to take advantage of any new Boost Test related features. However it doesn't look like there is anything substantial in that regard.

11. 
    DrahtBot commented at 3:33 PM on June 23, 2021: contributor

    <!--e57a25ab6845829454e8d69fc972939a-->

    The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

    <!--174a7506f384e20aa4161008e828411d-->

    Conflicts

    Reviewers, this pull request conflicts with the following ones:

    • #22294 by laanwj
    • #20744 by fanquake
    • #19460 by ryanofsky
    • #10102 by ryanofsky

    If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

12. DrahtBot cross-referenced this on Jun 23, 2021 from issue build: Perform boost::process detection with BOOST_CPPFLAGS by laanwj
13. DrahtBot cross-referenced this on Jun 24, 2021 from issue Multiprocess bitcoin by ryanofsky
14. 
    laanwj commented at 8:43 AM on June 24, 2021: member

    Tested ACK 957f3584277d33b44b9f1618142f743e4083eb5d that this fixes #22269. No opinion on versions.

15. 
    MarcoFalke commented at 8:55 AM on June 24, 2021: member

    review ACK 957f3584277d33b44b9f1618142f743e4083eb5d

    didn't test

16. MarcoFalke added this to the milestone 22.0 on Jun 24, 2021
17. fanquake merged this on Jun 24, 2021
18. fanquake closed this on Jun 24, 2021

    ---

19. fanquake deleted the branch on Jun 24, 2021
20. sidhujag referenced this in commit 82f2101bf7 on Jun 24, 2021
21. 
    luke-jr commented at 12:42 AM on June 27, 2021: member

    Please check at least RHEL next time. For reference, looks like it's got 1.66 so should be fine here.

22. 
    fanquake commented at 2:14 AM on June 28, 2021: member

    Please check at least RHEL next time. For reference, looks like it's got 1.66 so should be fine here.

    This was somewhat covered by the fact that RHEL is based off a semi-recent Fedora. It would be fine in any case, because if it had a too-old version of Boost, anyone wanting to compile on that system can use depends.

23. hebasto cross-referenced this on Aug 19, 2021 from issue bump boost test to 1.59? by MarcoFalke
24. barton2526 cross-referenced this on Sep 8, 2021 from issue refactor: Fix advanced compiler warnings by jamescowens
25. mjdietzx cross-referenced this on Oct 29, 2021 from issue refactor: get wallet path relative to wallet_dir by mjdietzx
26. bitcoin deleted a comment on Oct 30, 2021
27. bitcoin deleted a comment on Oct 30, 2021
28. fanquake referenced this in commit 7efc628539 on Oct 30, 2021
29. sidhujag referenced this in commit 7704bb722b on Oct 30, 2021
30. hebasto cross-referenced this on Mar 20, 2022 from issue main.cpp error during make. Ubuntu 16.04 with boost 1.77 by cvcook7
31. gwillen referenced this in commit 91ecc65c31 on Jun 1, 2022
32. bitcoin locked this on Oct 30, 2022

Contributors

fanquakeMarcoFalkehebastoDrahtBotlaanwjluke-jr

Milestone
22.0