boost 1.68.0

[lembacon]

• Have you followed the guidelines for contributing?
• Have you checked that there aren't other open pull requests for the same formula update/change?
• Have you built your formula locally with brew install --build-from-source <formula>, where <formula> is the name of the formula you're submitting?
• Does your build pass brew audit --strict <formula> (after doing brew install <formula>)?

---

boost-build has not been updated to 1.68.0 yet. https://github.com/boostorg/build/releases

Edited: boost-build has been updated to 1.68.0.

[eyal0]

@lembacon Hello!

I am a maintainer of pcb2gcode. pcb2gcode uses boost geometry heavily.

Boost geometry 1.67 and up is severely broken: https://svn.boost.org/trac10/ticket/13645 No one has responded to my issue.

boost 1.66 is the latest working boost that I have found. For my OSX users, this is a problem because there are not versions between boost 1.60 and 1.67 available on brew.

I do not have any OSX computers or experience. Can you or someone else add boost 1.66 to homebrew? Or could you point me to someone that can?

My users would appreciate this. Thanks!

[lembacon]

Since @DomT4 pointed out that

[…], it’s worth splitting anything that breaks into individual PRs and handling those after the main boost update has been pushed through successfully. Trying to get a boost update done in one PR generally results in it taking weeks.

It will indeed take us forever to update boost, so I'm temporarily dropping all commits regarding revision bumps, and will fix test failures due to broken linkages in a separated PR.

[Edited] Actually there should be plenty of separated PRs though. 🤨

[lembacon]

@eyal0 I'm take a look later. Anyway, a pull request is always welcome though. :-)

[DomT4]

It will indeed take us forever to update boost, so I'm temporarily dropping all commits regarding revision bumps

It's worth running it through once, leaving anything in that works or can be fixed in a trivial way, but pulling out anything that is broken in that initial run into a unique PR. It's possible not everything needs a bump; last time around @commitay only updated ~30 formulae. That number is pretty consistent so I think it's worth starting with that list and see what happens.

boost 1.66 is the latest working boost that I have found. For my OSX users, this is a problem because there are not versions between boost 1.60 and 1.67 available on brew.

It's likely one would need to 💀 if we added a new one. We already have more old boost versions sitting around than we should.

[lembacon]

Pushed revision bumps for 29 formulae. 🤨

[lembacon]

Some updates here after an, unsurprisingly, failed CI run:

• Pushed a few more revision bumps.
• Removed obsolete -mt suffix for ncmpcpp.
• Applied upstream patch for sfcgal, see Find CGAL before Boost to prevent it overwriting Boost_LIBRARIES. Oslandia/SFCGAL#169.
• Applied upstream patch for osquery, see Unbreak build against Boost 1.68 osquery/osquery#4692.
• mkvtoolnix and osrm-backend are both still failing due to Undefined symbols […]: […]boost::system::detail::system_category_instance.
• pcl and ledger are failing due to fatal error: 'boost/uuid/sha1.hpp' file not found.

[eyal0]

@lembacon No, I think that I figured out what is happening and the problem was mine.

First, bg:simplify on a polygon now changes the start/end point of a ring. That is a new behavior that surprised me but it's correct.

Second, passing a vector of pairs of doubles to bg::simplify used to simplify it like a linestring. That is no longer true. The behavior is strange now. But if my input is a valid input with the right geometry concept, it works fine.

So my problem went away when I fixed my input type. No need for boost 1.66 for me anymore.

Thanks for your hard work!

[lembacon]

@eyal0 So we can drop the vendored boost 1.66.0 when a new release of pcb2gcode is available. 😉

[jbeich]

mkvtoolnix and osrm-backend are both still failing due to Undefined symbols ... system_category_instance

Try relplacing -std=c++11 with -std=c++14 per boostorg/system#24
https://github.com/Homebrew/homebrew-core/blob/c7893f954fefda5076144a161014e349336c742a/Formula/boost.rb#L74

[lembacon]

Try relplacing -std=c++11 with -std=c++14 per boostorg/system#24

If we decide to bump the C++ standard used to build Boost from C++11 to C++14, then we may really need to rebuild everything depending on Boost. At least, this is something worthing considering. As of LLVM 6.0.0, which was released on 8 Mar 2018, the default C++ standard is C++14, IIRC. cc @Homebrew/maintainers

[fxcoudert]

Let's try not to move C++ standard to much… it might break other stuff :(

[DomT4]

@BrewTestBot test this please

[DomT4]

It's fairly quiet right now so we'll try and get a build through. Go ahead and kill it though if CI gets super busy again.

[lembacon]

The CI is started again here, feel free to kill it.

[lembacon]

Wish I could add a versioned boost@1.67 for those failed formulae. 🤒

[DomT4]

Hopefully it'll be okay to get through 🤞. It's been a quiet day for CI thus far.

[fxcoudert]

@BrewTestBot test this please

[JoeLoser]

Is there a decision on how to move forward with this PR? I'd love to see this get into master so I do not keep maintaining a Boost 1.68 built from source rather than Homebrew.

[claui]

@BrewTestBot test this please

[lembacon]

@JoeLoser I didn't find myself some time to handle this as well as #31193, I'm happy to close these PRs in favor of yours, if you do want to take them over. :-)

[lembacon]

Superseded by #32610.