Hey. I think I encoutered a bug that makes graphit...
# report-bugs
f
Hey. I think I encoutered a bug that makes graphite basically useless for me. I had a stack for multiple PRs. the frist one got merged. Now I run
gt sync
- and switched to a different branch. Then tried to switch back with
gt co BRANCH
- told me couldn't find the branch locally (whcih is wierd in the first place) - then I used
gt get BRANCH
to get it (that worked) - another
gt sync
is also running fine. But I did a linting change on a single file now. run
gt modify
(worked fine) but
gt submit
fails with a pre-submit 500 error
p
(If applicable) To help us repro, please send the following if you haven't already: - PR links, screenshots, and/or debug output if applicable - Repro steps - Does this happen consistently? Just to you or others on your team also? - Is this a regression? Is the problem new? - If you're having an issue with the CLI, please send us debug logs using
gt feedback --with-debug-context <any message>
f
I already submitted feedback this morning, but never got a response. Also the repo was not recentely renamed, which is the only mention of that error I could find online
So basically I am now blocked from updating my code
Something is def. broken in graphite here. I just recloned the repo. and used
gt get
for the branch. It is displayed in graphite like this:
What CLI shows me is this:
p
hey Felix, looking into this. We did also receive your prior reports- they were overnight for us here in the US.
👍 1
f
I see thanks. For now I just went around and used git push now. But it seems to me that something in the stacking or tracking logic got broken there
DM me should you need more details
@Gareth Brickman we have run into this again now. It would be great to have some sort of support on this?
p
hey Felix, what's the present issue?
f
We are getting pre-submit 500 errors on various of our stacks. We then workaround by just using
git push
but after 1 or 2 prs in teh stack the pop up again
p
gotcha, when one of those 500 errors occurs please run
gt feedback --with-debug-context <any message>
to share logs and info about your particular
gt
CLI with us. We may be able to find an additional message or request ID to investigate further on our end
f
Okay will do
Just run into it again and ran gt feedback.
As workaroudn we now need to run
git pull
then
git push
to get it back on track. It seems this happens when conflicts arise, but for some reason they are not tracked on
gt sync
p
hey Felix, received the
gt
feedback but it just contained the text message and no logs. Was it run with the
-with-debug-context
flag? That's necessary to get those to us
f
No wasn't. We just ran into this again, so it next we submit will be with debug context
Should be submitted
It appears this is related to conflicts in the branches.
This time we had to: git pull
resolve conflict
git add ... (files) git commit gt restack
resolve conflict AGAIN in gt
gt continue gt submit Really annoying. Then tried to submit feedback again (so you have after picture) - but ran into a 500 error from gt feedback as well. (tried multiple times) - hard to justify this for prod workload
p
gotcha, are multiple users collaborating on the same branch/stack? There's some more nuance to doing that with a recommended workflow of commands we document here
f
Not the same branch. Potentially same stack.
Also read through this - yes these are the commands we are following
p
gotcha, there's known frictions to collaboration when merge conflicts occur that the team's working to improve in
gt
and intend to roll out in the near future. Here's some additional guidance, which we'll be extending the existing stack collaboration doc with: 1. Don’t make meaningful changes (restacking is generally ok) to to PRs that you didn’t create. a. Calling
gt sync
before
gt submit
is one way to guard against this. 1. It follows that if you see a merge conflict unrelated to your changes, wait for the author of those changes resolve them first, since they have more context. If you want to resolve them yourself, then let them know. 2. Communicate if you are going to make changes that could affect someone else. 3. Only one person should work on a branch at a time.
f
Yeah. Overall this just generates friction in our deploy process. So most likely solution for us is to abandon graphite.
p
Avoiding toe-stepping during collaboration is definitely on our radar to address so I hope we can provide a much better experience for that imminently. In the meantime, I'm referring the error logs to our CLI team to look into in case there's some other underlying cause we can address, or at least communicate about more effectively in the CLI error response.
👍 1
f
Happened again. Few things to highlight here: • gt feedback takes looong to execute (not sure if it succeeded - assume not as it also has a 500 response) • Error (500) is very non-descriptive and indicates non-user error to me
image.png
p
the feedback submit did end up succeeding, it just had to be chunked due to the size of the logs being shared. Agree that's not a good user experience, I'll let the CLI team know that's happening
f
Also in my workarounds I ended up doing: • git push • gt sync • gt restack But then it ended up in conflict again after next changed (gt modify and gt submit - resulted in 500 again)
It would be helpful to know if there is a recommended strategy.
Maybe a command like
gt repair
would help
Even if that requires a few conflicts that need manual resolving
p
When you run
gt sync
are you alerted to a conflict? (That should automatically restack all branches that can be without conflicts)
And do you know if another user is pushing to this branch/PR around the same time as you are?
f
No I am not.
No. We are a small team. But we have working on stacked changes.
So while another user might not push to branch their gt submit might (if they update all branches in stack)
But not sure that is what happened here. As for this specific case I should have been the only one working on it (as it is the lowest branch in stack)
Also is there a timeline on the improvements for toe-stepping
p
I've been in the design discussions with the CLI team and they should start work on them imminently
❤️ 1
f
Still running into issues there. Currently I have one branch were it persistently changes back a part of the code I never touched to a version 2 items ago in the main. Which then generates conflicts. Super annoying to fix every time