:Gina commit fails in repo with multiple remotes

[ulidtko]

Hi! First of all, nice plugin, enjoying it, thanks a lot.

Now, the "usual" simple commit works just fine for me from Gina. However I've just needed to do funny stuff with git remotes, and consequently get this on :Gina commit:

[gina] ['.git/hooks/prepare-commit-msg', '.git/COMMIT_EDITMSG', 'message']                                                                                                      
[gina] /bin/sh: 2: local-main2/master-4.4..snmp-agent: not found
[gina] Traceback (most recent call last):
[gina]   File ".git/hooks/commit-msg", line 37, in <module>
[gina]     shell=True).decode('utf-8').split()
[gina]   File "/usr/lib/python3.6/subprocess.py", line 336, in check_output
[gina]     **kwargs).stdout
[gina]   File "/usr/lib/python3.6/subprocess.py", line 418, in run
[gina]     output=stdout, stderr=stderr)
[gina] subprocess.CalledProcessError: Command 'git rev-list --no-merges gitlab
[gina] local-main2/master-4.4..snmp-agent
[gina] ' returned non-zero exit status 127.
Press ENTER or type command to continue

It looks to me as if the git rev-list cmdline has got embedded newlines in it.

I'll explain the "funny stuff" with git remotes next — assuming that the reader understands the basic concepts of git remotes.

TL;DR: if my repo has two remotes — :Gina commit doesn't work.

---

We use a private GitLab instance. It has a repository called main. I also have two local clones on my machine, ~/src/main and ~/src/main2. Both of them have the default post-clone remote renamed origin → gitlab. An example of a remote branch is then gitlab/master-4.4; with this I can do things like git rebase gitlab/master-4.4. So far so good.

Now, I'm not a huge fan of keeping two separate clones around, but sometimes that comes as a real time-saver. Long story short, I ended up with commits in main2 repo which I needed to pull into main. Those commits are the branch snmp-agent.

Git is perfectly capable to treat a "neighbour" clone as a remote (that's what the file:// remote URIs essentially are). So I did:

$ cd ~/src/main
[main]$ git remote add local-main2 ../main2
[main]$ git fetch local-main2

If you follow me, this output would appear just as expected:

[main]$ git remote -v
gitlab	[email protected]:ipmp/main.git (fetch)
gitlab	[email protected]:ipmp/main.git (push)
local-main2	../ipmp-main2 (fetch)
local-main2	../ipmp-main2 (push)

Two remotes, alright? One gitlab, one local-main2.

---

Next, I get to pinpoint my work made in repo main2 as a branch snmp-agent in repo main:

[main]$ git checkout -b snmp-agent local-main2/snmp-agent
[main]$ git branch -u gitlab/master-4.6
[main]$ git status
On branch snmp-agent
Your branch and 'gitlab/master-4.6' have diverged,
and have 41 and 495 different commits each, respectively.
  (use "git pull" to merge the remote branch into yours)

It's been a while, so I replay the changes on top of the updated gitlab/master-4.6, just as I always do.

[main]$ git pull --rebase
[...]
[main]$ git status
On branch snmp-agent
Your branch is ahead of 'gitlab/master-4.6' by 41 commits.
  (use "git push" to publish your local commits)

Great! Here, the "funny git stuff" ends. So I went back into vim, made some changes, invoked :Gina status, staged the changes and proceeded with :Gina commit, again as usual. Upon filling out the commit message, however I see the traceback as above (and no commit happens, and the long long commit message is lost):

[gina] ['.git/hooks/prepare-commit-msg', '.git/COMMIT_EDITMSG', 'message']                                                                                                     
[gina] /bin/sh: 2: local-main2/master-4.4..snmp-agent: not found
[gina] Traceback (most recent call last):
[gina]   File ".git/hooks/commit-msg", line 37, in <module>
[gina]     shell=True).decode('utf-8').split()
[gina]   File "/usr/lib/python3.6/subprocess.py", line 336, in check_output
[gina]     **kwargs).stdout
[gina]   File "/usr/lib/python3.6/subprocess.py", line 418, in run
[gina]     output=stdout, stderr=stderr)
[gina] subprocess.CalledProcessError: Command 'git rev-list --no-merges gitlab
[gina] local-main2/master-4.4..snmp-agent
[gina] ' returned non-zero exit status 127.

That local-main2/master-4.4 looks really weird in the traceback. It shouldn't appear there at all — as I've definitely switched the upstream of snmp-agent to gitlab/master-4.6. The git rev-list invokation is totally borked. From the /bin/sh: ...: not found stderr, I deduce that the last three lines are not wrapping artifact, but two newlines embedded in git rev-list cmdline.

---

I hope, you can already guess what's wrong with Gina source 😃

:Gina commit works again as soon as I do git remote remove local-main2.

[lambdalisue]

Thank you for the detail 👍 It sounds a bug and I'll check 👍

[lambdalisue]

Hum... I tried to reproduce but it's too complicated. Could you make the minimum case?

[ulidtko]

Ohhh... I'm sorry!

I tried minimizing the test, and quickly found out that Gina has nothing to do with this -- it was a local commit hook (.git/hooks/commit-msg) failing, in one specific repo. I just misinterpreted the origin of that stacktrace 😄

Thanks again; closing.