[RFC] Add dependency rebuilding (32/64 bit).

[fwalch]

This has been used to generate neovim/deps#4. Here's a full Travis test build: https://travis-ci.org/fwalch/bot-ci/builds/35124444. Output is in https://github.com/fwalch/deps/tree/testing.

[justinmk]

small nit: travis-deps.sh ? Nevermind, I see why you named it like this.

[fwalch]

Changed the file names; build is running at https://travis-ci.org/fwalch/bot-ci/builds/35124444.

[fwalch]

Added a prompt to push to a repository for local builds and adapted the README. Example commit from a local build: fwalch/neovim-deps@d7dd51b

[justinmk]

I think ci/common/neovim.sh, ci/common/doc.sh, and ci/common/{dependencies.sh names should not be changed. Only deps-repo.sh is a special case to avoid confusion; but changing the others adds confusion.

It's not a huge concern, but I would like to make it easy for other devs to keep up with the changes in bot-ci.

[fwalch]

Hm, my idea was that changing all repository-related library scripts to -repo would increase consistency and thus understandability. But if you think otherwise, I'll change it. For the Travis dependencies, should I keep travis-deps.sh or rename it back to dependencies.sh?

[justinmk]

Yeah, but it's not really clear what neovim-repo really means, for example. We need neovim.sh for a number things; yes it comes from a repo, but that isn't really relevant.

rename it back to dependencies.sh?

I think it should stay dependencies.sh.

[fwalch]

Okay, done. Also, if /opt/neovim-deps is an accessible Git repo, the script now won't delete the directory, but pull from the given DEPS_REPO/DEPS_BRANCH. To not interfere with the existing repo too much (change Git config, write to .git/info/sparse-checkout), it won't use Git subtree for local builds.

Build is at https://travis-ci.org/fwalch/bot-ci/builds/35143356.

[justinmk]

Very slick :)

[justinmk]

@fwalch I just ran bot-ci and it failed on a few things. I won't be able to investigate for the next few hours. If you have a chance to look at the logs, any idea if this PR will fix it?

[fwalch]

@justinmk The recent directory changes (/32/, /64/) to neovim/deps are causing this, I think.. so this PR should fix it. The latest test build ran successfully, and that was based on neovim/neovim@6a8932a (so using an up-to-date deps repo).

[fwalch]

@justinmk Updated with is_ci_build, successful build at https://travis-ci.org/fwalch/bot-ci/builds/35167250.

[justinmk]

Won't this return 0/truthy for non-local builds?

[fwalch]

Yes, this is for things that the user might want to abort locally, but Travis should do in all cases (e.g. https://github.com/neovim/bot-ci/pull/20/files#diff-eb23d815554bddb9ac4f5c48c7f8601fR60).

[justinmk]

Thanks again!

[fwalch]

@justinmk Oh, infinite push loop:

Retry pushing to neovim/deps master.
From git://github.com/neovim/deps

• branch master -> FETCH_HEAD
  Current branch master is up to date.

You have about 55 minutes (until Travis aborts/times out the build) to give Marvim write access to neovim/deps 😁

[justinmk]

Thanks for catching that!

[justinmk]

@tarruda @fwalch The vroom tests are failing with this:

No handlers could be found for logger "neovim.uv_stream"

�[3g        �H        �H        �H        �H        �H        �H        �H        �H        �H
�c�[!p�[?3;4l�[4l�>
�[1m�[36mtest/legacy/test2-filename-recognition.vroom�(B�[m

�[33mERROR�(B�[m: Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/vroom/runner.py", line 70, in __call__
    self.env.vim.Start()
  File "/usr/local/lib/python2.7/dist-packages/vroom/neovim_mod.py", line 45, in Start
    self.conn = neovim.connect(self.args.servername)
  File "/usr/local/lib/python2.7/dist-packages/neovim/__init__.py", line 22, in connect
    client.discover_api()
  File "/usr/local/lib/python2.7/dist-packages/neovim/client.py", line 255, in discover_api
    channel_id, api = self.rpc_request("vim_get_api_info", [])
  File "/usr/local/lib/python2.7/dist-packages/neovim/client.py", line 92, in rpc_request
    err, result = self.rpc_blocking_request(method, args)
  File "/usr/local/lib/python2.7/dist-packages/neovim/client.py", line 71, in rpc_blocking_request
    self.stream.send(method, args, response_cb)
  File "/usr/local/lib/python2.7/dist-packages/neovim/rpc_stream.py", line 32, in send
    self.stream.send([0, request_id, method, args])
  File "/usr/local/lib/python2.7/dist-packages/neovim/msgpack_stream.py", line 36, in send
    self.stream.send(self.packer.pack(msg))
  File "/usr/local/lib/python2.7/dist-packages/neovim/uv_stream.py", line 170, in send
    raise err
IOError: no such file or directory

Seems like a path might need to be updated in the python-client or in the vroom task for travis?

[fwalch]

Hm, strange.. I ran a build from the deps PR, and it worked: https://travis-ci.org/fwalch/neovim/builds/35042442. I only found paths in .travis.yml (neovim/pynvim#30).

[justinmk]

I ran a build from the deps PR, and it worked: https://travis-ci.org/fwalch/neovim/builds/35042442. I

@fwalch Can you restart that build and see what happens? Something between yesterday and today broke something, somehow. If that same build succeeds, we can at least rule out the deps repo I guess. Nevermind, just noticed it's running against a different branch: fwalch/neovim@14c6cb2

[fwalch]

When I exit nvim, sometimes I get the following message:

Error detected while processing /usr/share/nvim/runtime/plugin/python_setup.vim:
line   41:
Before returning from a RPC call, channel 1 was closed by the client
Press ENTER or type command to continue

Could it be that it fails trying to log/output this error message (error occurs at raise err)?

[justinmk]

Good question! I don't see that error, but it could be a clue.