Workflow for building on macOS

[jwodder]

This is (for now) just a draft created in order to trigger a build and make sure things are working.

[yarikoptic]

Wow -- you "translated" all test scenarios. the "crippled-home" seems stuck -- running for 3h and no output. I think I had a similar problem before: https://github.community/t/no-log-at-all-online-or-in-bundled-download-for-the-last-step-in-the-action/18527 which I forgot details of the resolution, so it might indeed have been what I stated in the last comment there. I also remember that OSX build of git-annex had issues with "too many open files" but why would it be on crippled FS?

For now (or forever) IMHO it is ok to kick out crippled-home test scenario, unless you could easily replicate locally on your laptop and report back. Unlikely (I hope!) any user on OSX will have their home on some crippled FS.

[yarikoptic]

left a comment to add version info

but I also wonder what in the longer run would be the best way to reuse step definitions across workflows? we could use jinja2 for these as well as we do for testing extensions, but may be there is a better way now?

[yarikoptic]

Please add here a step which would print the version since ATM we follow original build close and git-annex.dmg does not include version in its filename (unlike what we have for .deb packages, so it is possible to discover version really easy). FWIW -- we could have that step in both linux and osx workflow. git annex version output could be generally useful since summarizes build params etc

[jwodder]

Done.

[yarikoptic]

I have canceled it and

log appeared -- seems to be a legit issue -- got stuck!

Detected a filesystem without fifo support.
  Disabling ssh connection caching.
  Filesystem allows writing to files whose write bit is not set.
  Detected a crippled filesystem.
  Disabling core.symlinks.
  Entering an adjusted branch where files are unlocked as this filesystem does not support locked files.
Switched to branch 'adjusted/master(unlocked)'
git-annex: openFile: resource busy (file is locked)
      init:                                               FAIL (2.18s)
        Test.hs:249:
        init failed
      unavailable remote
##[error]The operation was canceled.

reported against git-annex

[jwodder]

For now (or forever) IMHO it is ok to kick out crippled-home test scenario, unless you could easily replicate locally on your laptop and report back.

I can replicate the problem on my computer; using a FAT32 image as $HOME causes git annex test to hang indefinitely a couple seconds in. I don't know why it's happening, though.

[yarikoptic]

just remove crippled-home from this list with a comment and url to git annex issue to readd it whenever joey fixes it. I need to do that for the NFS testing I had added recently but had to disable: https://github.com/datalad/datalad-extensions/blob/master/.github/workflows/build-git-annex-debianstandalone.yaml#L60 That issue is fixed in annex but build is failing -- yet to finish troubleshooting in #32

[jwodder]

with a comment and url to git annex issue

URL to what issue? #30? This PR? Something else?

[yarikoptic]

git annex issue: https://git-annex.branchable.com/bugs/gets_stuck_in_Remote_Tests___40____63____41___while_running_on_OSX_with_HOME_on_crippled_FS/?updated since that is the one, when fixed, would let us add crippled-home back to the matrix

[yarikoptic]

ssh setup for testing didn't work out: setup-docker-ssh: line 74: docker: command not found - I guess docker to be installed . Make script fail (set -eu) as soon as some command (or there is undefined variable - u portion) within it fails.

[jwodder]

ssh setup for testing didn't work out: setup-docker-ssh: line 74: docker: command not found - I guess docker to be installed . Make script fail (set -eu) as soon as some command (or there is undefined variable - u portion) within it fails.

The setup-docker-ssh script is from https://github.com/datalad-tester/docker-ssh-target. It might be better to remove the "master" and "maint" versions from the test-datalad matrix, as those are the ones that use Docker.

[yarikoptic]

we have switched to use dockerized setup for ssh testing recently in datalad. Why to remove them and not just to install docker ?

[jwodder]

Never mind. I've added a step that I think should install Docker, and I've created a PR for adding "set -eu" to the appropriate script: datalad/datalad#4920

[yarikoptic]

sorry - I have merged the set -eu in maint only for now... it will get to master eventually. no need to change this up ATM.

[yarikoptic]

superseded by #33