Use origin/branch when checking for new commits + python rewrite #16
Conversation
- Check v:vim_did_enter and call for debugging convenience
There was a lot of strange behavior in vimscript version of this plugin. The on_stdout callback appeared to be called a variable number of times before the on_exit callback, most likely indicating a race. I tried refactoring to call the rev-list command sequentially for each plugin in an async cascade, but for unknown reasons, stdout was blank for about half of the calls in my setup. In python, we can use a comparatively simple synchronous function wrapped in vim.async_call, so that's what I've done. The plugin now appears to finally work consistently and as expected. Since async/job.vim is no longer needed, I've removed that as well.
vim.eval returns a string, whereas we need an int for boolean testing
|
Oh, I also added a new option g:outdated_plugins_trigger_mode to trigger :PlugUpdate as needed |
Was using python2 subprocess.Popen, which accepts the text argument. Since python3 is utf8, this arg was removed (always returns utf8). Also got rid of unused stderr var to quiet lint
Got it backwards in the previous commit. subprocess always returns binary output now, so it must be decoded. Also used subprocess.run instead of Popen/communicate for simplicity
|
hi @thisisrandy! Thanks for your work and sorry for my late reply. I've tried it and it works much better than vimscript version. Good job! What is the minimal version of python required? I want to update the README file. |
|
It seems like I'm not able to add commits to this PR (I can't push to master branch on your forked repo). |
|
Thanks for your reply, @semanser. I've added you as a collaborator on my fork, so you should have push access now. I developed the python bit using 3.6.8, so I know that's safe. Probably should have added the future imports to make it backward compatible, though. I'll take care of that in a bit, or if you beat me to it, you're welcome to do the same. |
|
Actually, I take that back. The |
Note: this commit should be excluded from any pull requests, but it is useful for clarity should anyone want to use the fork directly
It's unclear when this broke, but g:plugs[name].branch is apparently now empty when not explicitly specified. Since this plugin never would have worked if it was always that way, presumably the behavior changed at some point
thisisrandy
force-pushed
the
master
branch
3 times, most recently
from
ca1dc76
to
4c27360
Compare
Don't guess branch if not specified explicitly
This fixes /issues/14 for me (I have coc.nvim checked out from the release branch).
Additionally, there was some other race-condition-like behavior going on where the on_stdout callback was apparently entered variably fewer times than expected (see 72c9dcf for details). I was unable to resolve it in vimscript, but a python rewrite did the trick.