🔧 avoid the installation of boost cmake config. by daminetreg · Pull Request #140 · cpp-pm/hunter

daminetreg · 2020-01-17T15:54:55Z

I've followed this guide
step by step carefully. [Yes]
I've tested this package remotely and have excluded all broken builds.
Here is the links to the Travis/AppVeyor with status "All passed 😉":
- https://ci.appveyor.com/project/daminetreg/hunter/builds/30255155
- https://travis-ci.com/nxxm/hunter/builds/145347045

This fixes the issues discovered in #133 : when linking to a Boost library, the Boost::boost INTERFACE lib gets to link to the /Source directory that hunters generate during the build.

This is due of the BoostConfig-*.cmake files generated by Boost, because they aren't cleaned up anymore by your script : @HUNTER_SELF@/scripts/clean-boost-configs.cmake. But we don't need it anymore as we can pass b2 --no-cmake-config.

The cleanup didn't work anymore is due to a change in the names, at least on macOS I could check that, that the boost b2 generated cmake package configs were named lowercase like boost_headers-1.72.0/boost_headers-config.cmake, and I think at least in some case takes precedence over the cmake module for finding boost.

Presumably that's the reason why tests didn't fail when @dan-42 updated the boost version, or there is some cmakecache thing or ci cache thing involved.

Little question should I update the Boost revision to avoid any successful build because of a cache that doesn't notice the changes ?

daminetreg · 2020-01-17T21:23:13Z

All tests passed for pkg.boost. ✔️

dan-42

👍

BenWhetton · 2020-01-20T18:00:32Z

Does the ID value on the last line of cmake/projects/Boost/hunter.cmake need to be incremented before this is merged in?
See https://github.com/cpp-pm/hunter/blob/master/docs/reference/user-modules/hunter_download.rst#hunter_download

daminetreg · 2020-01-20T19:17:28Z

Does the ID value on the last line of cmake/projects/Boost/hunter.cmake need to be incremented before this is merged in?
See https://github.com/cpp-pm/hunter/blob/master/docs/reference/user-modules/hunter_download.rst#hunter_download

Yes thanks for pointing the specific id out, that was the bit I was asking. I’ll push that change to PACKAGE_INTERNAL_DEPS_ID.

Bjoe · 2020-01-21T09:23:34Z

Read also my #133 (comment) about Boost::boost ....

daminetreg · 2020-01-21T11:07:34Z

I upgraded Boost PACKAGE_INTERNAL_DEPS_ID as pointed out by @BenWhetton. And relaunched tests on the test.boost branch.

See builds in progress :

Read also my #133 (comment) about Boost::boost ....

I've replied there, I think you are right on the long term, but as of Boost 1.72.0 I don't see another option as this one for now.

daminetreg · 2020-01-21T17:25:44Z

Again, with the PACKAGE_INTERNAL_DEPS_ID updated all tests passed for test.boost. ✔️

dan-42 · 2020-01-21T18:38:04Z

PING: @bkotzz @rbsheth 👋

Bjoe · 2020-01-23T05:30:51Z

Btw. I would say to be consistent, you should add the --no-cmake-config also in the iOS build see https://github.com/nxxm/hunter/blob/27532ea5f4093ba312bc53692553a363966a62f6/cmake/projects/Boost/schemes/url_sha1_boost_ios_library.cmake.in#L126

And what about the other PACKAGE_INTERNAL_DEPS_ID in for example https://github.com/nxxm/hunter/blob/27532ea5f4093ba312bc53692553a363966a62f6/cmake/projects/Boost/atomic/hunter.cmake#L21 ?

BenWhetton · 2020-01-27T09:17:29Z

@rbsheth: Will you be able to review this soon? We are waiting for this fix to land but if it is going to take much longer we'll have to implement a local workaround for the issues we're encountering (see #141).

rbsheth

@daminetreg Thanks for the detailed investigation and fix!

@BenWhetton Yes. 😄

I think the PACKAGE_INTERNAL_DEPS_ID needs to be updated here and the generate.sh script run.
https://github.com/cpp-pm/hunter/blob/master/cmake/projects/Boost/hunter.cmake.in#L21

As @Bjoe said, the iOS build may need to be updated with the --no-cmake-config also. See https://github.com/cpp-pm/hunter/blob/master/cmake/projects/Boost/schemes/url_sha1_boost_ios_library.cmake.in

Will adding the flag --no-cmake-config break older Boost versions that don't support the flag?

Bjoe · 2020-02-04T13:26:27Z

@rbsheth We should update PACKAGE_INTERNAL_DEPS_ID also in all boost components.

If we not do this, then every hunter user they already build boost 1.72.0 from hunter version 0.23.245 must nuked ~/.hunter directory, because for hunter the boost version is not changed so they will not build new boost package .... see also the comment/post in issue #141

BenWhetton · 2020-02-05T09:47:59Z

@daminetreg do you think you'll have time to update this soon?

As @Bjoe said, the iOS build may need to be updated with the --no-cmake-config also. See https://github.com/cpp-pm/hunter/blob/master/cmake/projects/Boost/schemes/url_sha1_boost_ios_library.cmake.in
Will adding the flag --no-cmake-config break older Boost versions that don't support the flag?

Are there any regression tests which cover this?

daminetreg · 2020-02-05T14:31:50Z

@daminetreg do you think you'll have time to update this soon?

Yes I'll do that this Friday morning. I have a little bit of time.

As @Bjoe said, the iOS build may need to be updated with the --no-cmake-config also. See https://github.com/cpp-pm/hunter/blob/master/cmake/projects/Boost/schemes/url_sha1_boost_ios_library.cmake.in
Will adding the flag --no-cmake-config break older Boost versions that don't support the flag?

Are there any regression tests which cover this?

Well if you want previous version to continue to work without this I need to restore the files responsible of deleting BoostConfig.cmake. I'll do something based on the version of boost if it's greater or equal 1.72 I'll do what I have here and otherwise go back.

rbsheth · 2020-02-12T17:59:17Z

@daminetreg Did you have an opportunity to fix this?

madmongo1 · 2020-02-13T08:20:19Z

@rbsheth looks like @daminetreg is busy. This PR is pretty important. Boost is central to many c++ projects.
If you can offer guidance as to what remains to be done, I'd be happy to open a new PR which incorportates these changes and the final fix.

daminetreg · 2020-02-13T08:22:52Z

I did lag a bit, but it is almost done, this evening pr will be updated with the little bits missing to prevent hunter cache issues.

…d see previous commit) and unneeded clean-boost-configs as we can avoid that boost generates them.

…e the appropriate bjam switch as file name changed and will work better.

… to get the benefits of the patch.

… file.

daminetreg · 2020-02-14T10:27:46Z

IMHO I'm done with all comments from the review, and avoided changing behaviour for older boost version (tested also with an older one locally).
Here is the links to the Travis/AppVeyor with status "Build in Progress":
- https://ci.appveyor.com/project/daminetreg/hunter/builds/30799737
- https://travis-ci.com/nxxm/hunter/builds/148937865

daminetreg · 2020-02-14T10:30:59Z

For a summary:

Rebased on master to integrate the changes made in between
Updated the ios build
Implemented the use of bjam --no-cmake-configs only if version is >= 1.72.0.
Updated all PACKAGE_INTERNAL_DEPS_ID

madmongo1 · 2020-02-14T10:39:54Z

Thank you for taking the time to do this. I bow before your CMake expertise!

…ependent cmake generation and on cmake documented behaviour.

daminetreg · 2020-02-14T17:40:56Z

Thank you for taking the time to do this. I bow before your CMake expertise!

Thank you very much 😄

However the CI saw I did oversee the fact that the generator expression didn't work on all platform the same, and before all I used them in a context where they are not officially allowed by the documentation. So I removed them in favor of a bit more traditional approach.

This has the downside to try to clean always the boost config cmakes even when we don't install them, but it's safer as it is not allowed by the cmake documentation (even though this part worked everywhere due to the fact the file is generated by file(GENERATE OUTPUT ...).)

And to that while they came quite handy in this situations they are not thought for the kind of use I did, so better less clever but more solid.

Here is the new links to the Travis/AppVeyor with status "Build in Progress":
- https://ci.appveyor.com/project/daminetreg/hunter/builds/30809619
- https://travis-ci.com/nxxm/hunter/builds/149020624

rbsheth

Thanks @daminetreg! Will merge when green

cmake/projects/Boost/schemes/url_sha1_boost.cmake.in

daminetreg · 2020-02-14T22:49:48Z

Thank you very much for the review. Tests are all green and passed.

By the way where is @ruslo ?

rbsheth · 2020-02-14T22:54:27Z

@daminetreg Great! @ruslo has passed this project onto myself and @bkotzz - he is focusing on other things.

daminetreg · 2020-02-17T09:18:23Z

Thanks for merging ^^ Hope it will work for the users as CI confirmed.

@ruslo is a terribly great expert in Cmake, hope he can comeback someday ^^.

@rbsheth keep up with the super good work !

madmongo1 · 2020-02-17T09:57:45Z

confirmed working here:
https://github.com/madmongo1/dynamic_buffer_v2

daminetreg mentioned this pull request Jan 17, 2020

Boost 1.72.0-p0 fails on Boost::boost #133

Closed

daminetreg changed the title ~~🔧 avoid the installation of boost cmake config.~~ 🔧 avoid the installation of boost cmake config. Jan 18, 2020

daminetreg requested review from dan-42 and rbsheth January 18, 2020 10:10

dan-42 approved these changes Jan 20, 2020

View reviewed changes

daminetreg mentioned this pull request Jan 20, 2020

Boost library link dependencies are incorrect #141

Closed

rbsheth reviewed Jan 27, 2020

View reviewed changes

Bjoe mentioned this pull request Feb 13, 2020

Add support to use cmake package configuration file from boost #154

Merged

Damien Buhl added 2 commits February 13, 2020 23:21

🔧 avoid the installation of boost cmake config.

bef27f7

🔧 removes now no-more working (file names of boostconfig.cmake change…

02fa968

…d see previous commit) and unneeded clean-boost-configs as we can avoid that boost generates them.

daminetreg force-pushed the feature/fix-boost-1.72-cmake-package-configs branch from 27532ea to e7a71de Compare February 14, 2020 08:36

daminetreg added 2 commits February 14, 2020 11:04

🔧 For Boost older 1.72.0 keep deletion of boost configs, for newer us…

bc6be7b

…e the appropriate bjam switch as file name changed and will work better.

🔧 update PACKAGE_INTERNAL_DEPS_ID to avoid needing to nuke hunter dir…

b776b48

… to get the benefits of the patch.

daminetreg force-pushed the feature/fix-boost-1.72-cmake-package-configs branch from e7a71de to b776b48 Compare February 14, 2020 10:05

😬 roll back broken package warning after regeneration of each library…

2827848

… file.

🔧 Remove elegant cmake generator expression in favor of generator-ind…

3d9b33a

…ependent cmake generation and on cmake documented behaviour.

daminetreg requested review from dan-42 and rbsheth February 14, 2020 17:50

rbsheth approved these changes Feb 14, 2020

View reviewed changes

cmake/projects/Boost/schemes/url_sha1_boost.cmake.in Show resolved Hide resolved

rbsheth merged commit 67c40b9 into cpp-pm:master Feb 14, 2020

Comments

Conversation

daminetreg commented Jan 17, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

daminetreg commented Jan 17, 2020

Uh oh!

dan-42 left a comment

Choose a reason for hiding this comment

Uh oh!

BenWhetton commented Jan 20, 2020

Uh oh!

daminetreg commented Jan 20, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Bjoe commented Jan 21, 2020

Uh oh!

daminetreg commented Jan 21, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

daminetreg commented Jan 21, 2020

Uh oh!

dan-42 commented Jan 21, 2020

Uh oh!

Bjoe commented Jan 23, 2020

Uh oh!

BenWhetton commented Jan 27, 2020

Uh oh!

rbsheth left a comment

Choose a reason for hiding this comment

Uh oh!

Bjoe commented Feb 4, 2020

Uh oh!

BenWhetton commented Feb 5, 2020

Uh oh!

daminetreg commented Feb 5, 2020

Uh oh!

rbsheth commented Feb 12, 2020

Uh oh!

madmongo1 commented Feb 13, 2020

Uh oh!

daminetreg commented Feb 13, 2020

Uh oh!

daminetreg commented Feb 14, 2020

Uh oh!

daminetreg commented Feb 14, 2020

Uh oh!

madmongo1 commented Feb 14, 2020

Uh oh!

daminetreg commented Feb 14, 2020

Uh oh!

rbsheth left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

daminetreg commented Feb 14, 2020

Uh oh!

rbsheth commented Feb 14, 2020

Uh oh!

daminetreg commented Feb 17, 2020

Uh oh!

madmongo1 commented Feb 17, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

daminetreg commented Jan 17, 2020 •

edited

Loading

daminetreg commented Jan 20, 2020 •

edited

Loading

daminetreg commented Jan 21, 2020 •

edited

Loading