[Documentation] Workflow for updating a dependency using git

[schomatis]

I'm looking for documentation that would help me with the scenario described below which focuses on the workflow of using gx with git to keep track of changes.

I've installed go-ipfs and I want to work and update one of its dependencies go-ipfs-cmds using git, that is, being able to track down my work and push it to the corresponding repo. After downloading and installing all of go-ipfs dependencies (make install) I've proceeded to use the gx-go link tool to replace the plain go-ipfs-cmds gx version with the git version cloned from the repo (see #161 (comment)).

At this point the go-ipfs-cmds dependency is linked to the cloned git repo and all its dependencies (as reported by git) have been updated to use gx, e.g.,

-	"github.com/ipfs/go-ipfs-cmdkit"
+	"gx/ipfs/QmceUdzxkimdYsgtX733uNgzf1DLHyBKN6ehGSp85ayppM/go-ipfs-cmdkit"

This makes sense because I still want go-ipfs-cmds to use the same dependencies as go-ipfs (because I'm testing it through the use of go-ipfs).

I can now work on go-ipfs-cmds and make the necessary changes but when I want to commit those changes (and later push them) I'm going to need to undo the import path modifications, for that I'm guessing I need to use gx-go rw, gx-go rw --undo or gx-go uw (I haven't figured out which yet). After the commit if I want to keep working on the repo I'm going to have to rewrite the imports again (and restore them again to commit).

All this is a rather cumbersome series of steps and I think I might be missing something (a gx concept, a gx-go command) that would help me smooth this workflow but I haven't found much information in the gx/gx-go READMEs related to this.

[whyrusleeping]

gx-go uw is an alias for gx-go rw --undo, and is the opposite of gx-go rw.

After making your changes to the linked package, you'll need to do a gx-go uw before publishing.

All this is a rather cumbersome series of steps and I think I might be missing something.

For now, that's what we have. I'm very keen on getting input from users on what would make a better process, so let me know what you think we can do to improve things