Add github ci for alpine and static

[thibaultduponchelle]

Hello,

Some new CI jobs so it could alert quickly if a commit breaks a different libc or a different linking method.

Could be nice to have a macos job also, but I got various non trivial errors when I tried... that's why it's not part of this PR 😁
Feel free to reimplement in travis if you prefer, but to my eyes it looked like more like a big compiler variants and versions check and I felt not comfortable to add it to travis.

Regards.

Thibault

[thibaultduponchelle]

I just seen the 2 issues #293 and #291 and now I feel like a bit like if I "cross-posted" 🤣

FYI, my CI tests are here : https://github.com/thibaultduponchelle/gm/actions
(I created a new repo, not sure how to test github ci on a fork)

Static and alpine are green today and one week ago.

Concerning macos, last error that I have is :

CXX      libtest/libtest_la-cmdline.lo
In file included from bin/gearadmin.cc:84:
In file included from /usr/local/include/boost/program_options.hpp:15:
In file included from /usr/local/include/boost/program_options/options_description.hpp:13:
In file included from /usr/local/include/boost/program_options/value_semantic.hpp:12:
In file included from /usr/local/include/boost/any.hpp:20:
In file included from /usr/local/include/boost/type_index.hpp:29:
In file included from /usr/local/include/boost/type_index/stl_type_index.hpp:47:
In file included from /usr/local/include/boost/container_hash/hash.hpp:24:
In file included from /usr/local/include/boost/type_traits/is_enum.hpp:14:
In file included from /usr/local/include/boost/type_traits/intrinsics.hpp:16:
/usr/local/include/boost/type_traits/detail/config.hpp:85:52: error: '__clang_major___WORKAROUND_GUARD' is not defined, evaluates to 0 [-Werror,-Wundef]
      && !BOOST_WORKAROUND(BOOST_MSVC, < 1900) && !BOOST_WORKAROUND(__clang_major__, <= 4)
                                                   ^
/usr/local/include/boost/config/workaround.hpp:246:10: note: expanded from macro 'BOOST_WORKAROUND'
       ((symbol ## _WORKAROUND_GUARD + 0 == 0) &&     \
         ^
<scratch space>:71:1: note: expanded from here
__clang_major___WORKAROUND_GUARD
^
  CXX      libtest/libtest_la-collection.lo
1 error generated.
make[2]: *** [bin/gearadmin.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

Compilation was OK when we were using the release tarball instead sources.

[p-alik]

Thank you!
Just a thought:
Travis CI runs make test.
Maybe Github CI should run test also.

[esabol]

I personally would prefer to keep everything in Travis CI. Spreading things around to other CIs just seems like more things to check to me.

But I could probably be persuaded if everyone else thinks it's a good idea....

[esabol]

Also, these builds should be configured with --enable-ssl. All our Travis CI builds do that. Better code coverage.

[esabol]

That macOS bootstrapping issue is annoying. It's apparently a bug in the boostlib headers. I found some discussion and a possible workaround here:

boostorg/config#322

[p-alik]

I personally would prefer to keep everything in Travis CI. Spreading things around to other CIs just seems like more things to check to me.

But I could probably be persuaded if everyone else thinks it's a good idea....

Github Actions (GA) CI gives us a comfortable possibility to run tests in docker containers. If I would setup CI tests this days, I would prefer GA.

[esabol]

Github Actions (GA) CI gives us a comfortable possibility to run tests in docker containers. If I would setup CI tests this days, I would prefer GA.

But you can do Docker containers in Travis CI, so what’s there to gain by doing a hodgepodge of both?

[p-alik]

That macOS bootstrapping issue is annoying. It's apparently a bug in the boostlib headers. I found some discussion and a possible workaround here:

boostorg/config#322

Could anyone try to run MacOS build with current ax_compiler_version.m4, please?

[p-alik]

But you can do Docker containers in Travis CI, so what’s there to gain by doing a hodgepodge of both?

I haven't any experiences with Docker setup in Travis CI and obviously I got a wrong deduction from your comment here.

[thibaultduponchelle]

Hello,

You are right and I agree with you.

There is no good reason to use GA instead Travis.

I’m actually pushing for the idea of adding alpine/static/macOS ci but I’m not pushing for an implementation.

I will maybe try to implement in Travis but can’t promise it 😄

[thibaultduponchelle]

I looked at travis and seems like it does not support natively docker (still feasible with service: docker then running docker pull, docker run but we can't just give a container: alpine).

Then... I skip my turn for this task and I will decline this PR 😄 to leave it to someone bolder than me 😀

[esabol]

I looked at travis and seems like it does not support natively docker (still feasible with service: docker then running docker pull, docker run but we can't just give a container: alpine).

Sorry, why does that matter?

Here’s an example that uses the alpine Docker image in Travis CI:
https://medium.com/@russleyshaw/how-to-use-travis-ci-380c96c9a639

[p-alik]

FYI, my CI tests are here : https://github.com/thibaultduponchelle/gm/actions

@thibaultduponchelle, it looks like the repository has been removed from github. I'm looking for your github action setup for MacOS. Could you share this part of code, please?

[thibaultduponchelle]

Ah sorry I have it locally I think, will upload it that in a couple of hours

[thibaultduponchelle]

Sorry, even locally I only have alpine.yml and static.yml :/

[p-alik]

I would appreciate, if you could share both of them. I guess MacOS customization is feasible.

[thibaultduponchelle]

They are still viewable in this PR or I uploaded them here : alpine and static

[p-alik]

Thank you, @thibaultduponchelle. I've cloned your master branch.

[p-alik]

That macOS bootstrapping issue is annoying. It's apparently a bug in the boostlib headers. I found some discussion and a possible workaround here:
boostorg/config#322

By solving the bootstrapping issue, we'll run in an other once or more.

Could anyone try to run MacOS build with current ax_compiler_version.m4, please?

An update ax_compiler_version.m4 by serial 9 doesn't help.

[esabol]

May I suggest moving discussion of macOS bootstrapping problems to issue #291?