feat: update clones if they're present #61
Conversation
src/MultiDocumenter.jl
Outdated
| elseif isempty(current_branch) | ||
| @error "current_branch was empty" | ||
| end | ||
| run(`$(git()) -C $(doc.upstream) reset --hard origin/$(doc.branch)`) |
There was a problem hiding this comment.
I am also not super comfortable with the reset --hard here. Especially if doc.upstream is pointing into a subdirectory of a real directory. Actually, instead of -C, should probably use --git-dir..
There was a problem hiding this comment.
Is there ever any harm in hard-resetting a remote branch?
There was a problem hiding this comment.
I think if everything is normal, then it doesn't do anything wrong. But the cases I'm concerned about:
- Something has gone horribly wrong, and
doc.upstreamis not actually a separate Git repo anymore, and then this command nukes changes on your main repository. - The user has actually gone into the cloned directories and modified stuff, and then we nuke their updates.
So I am wondering if something like a git merge --ff-only would be the better option here. I didn't want to do a simple git pull, since that, I believe, would fetch all branches.
There was a problem hiding this comment.
What I meant is that reset --hard origin/mybranch can't ever nuke any local changes, because those are in mybranch instead.
There was a problem hiding this comment.
I might be mistaken, but I believe reset --hard will change the commit your current branch points to origina/mybranch (any commit not on other branches/repos will become lost-ish), and it will also nuke all uncommitted working tree changes.
There was a problem hiding this comment.
Oh, doh! Yeah, I just misread the command here :P
Ok, so how about we just checkout remote/mybranch for the build?
|
I updated this, making it Unrelatedly, I also added |
So if you are cloning the remotes to a local directory, it updates the local remotes. Not a 100% sure about this, but it would be good to have something like this.
It does slow things down a bit (not as bad as full clones, but still). So maybe we want to guard this with an environment variable or something?