Random crashes during interactive rebase

[ada4a]

Describe the bug
sometimes, when I start an interactive rebase, after 1-2 seconds everything suddenly dies. This is the error message that I get:

panic: runtime error: slice bounds out of range [1:0]

goroutine 3870 [running]:
github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers.(*BranchesHelper).AutoForwardBranches(0xc000147e00)
        github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers/branches_helper.go:277 +0x5fd
github.com/jesseduffield/lazygit/pkg/gui.(*BackgroundRoutineMgr).backgroundFetch(0xc000329c50)
        github.com/jesseduffield/lazygit/pkg/gui/background.go:131 +0x10a
github.com/jesseduffield/lazygit/pkg/gui.(*BackgroundRoutineMgr).startBackgroundFetch.func1.1({0xe89040?, 0xc003b77df8?})
        github.com/jesseduffield/lazygit/pkg/gui/background.go:80 +0x17
github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers.(*AppStatusHelper).WithWaitingStatusImpl.func1(0xc002299110?)
        github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers/app_status_helper.go:69 +0x53
github.com/jesseduffield/lazygit/pkg/gui/status.(*StatusManager).WithWaitingStatus(0xc00034cbd0, {0xf1ea57, 0x8}, 0xc0002c2450, 0xc003b77ea0)
        github.com/jesseduffield/lazygit/pkg/gui/status/status_manager.go:56 +0xd7
github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers.(*AppStatusHelper).WithWaitingStatusImpl(0xc000011380, {0xf1ea57, 0x8}, 0xc003b77f08, {0x0, 0x0})
        github.com/jesseduffield/lazygit/pkg/gui/controllers/helpers/app_status_helper.go:68 +0xc7
github.com/jesseduffield/lazygit/pkg/gui.(*BackgroundRoutineMgr).startBackgroundFetch.func1()
        github.com/jesseduffield/lazygit/pkg/gui/background.go:79 +0x5a
github.com/jesseduffield/lazygit/pkg/gui.(*BackgroundRoutineMgr).goEvery.func1.1({0x870734?, 0xa78933?})
        github.com/jesseduffield/lazygit/pkg/gui/background.go:112 +0x22
github.com/jesseduffield/gocui.(*Gui).onWorkerAux(0xc000218a80, 0x44e3b2?, {0x10e42c0?, 0xc002b3c020?})
        github.com/jesseduffield/gocui@v0.3.1-0.20250421160159-82c9aaeba2b9/gui.go:703 +0x65
github.com/jesseduffield/gocui.(*Gui).OnWorker.func1()
        github.com/jesseduffield/gocui@v0.3.1-0.20250421160159-82c9aaeba2b9/gui.go:690 +0x2b
created by github.com/jesseduffield/gocui.(*Gui).OnWorker in goroutine 275
        github.com/jesseduffield/gocui@v0.3.1-0.20250421160159-82c9aaeba2b9/gui.go:689 +0x90

[Process exited 2]

To Reproduce
Steps to reproduce the behavior:

1. start an interactive rebase (by pressing i)
2. (move a couple commits here and there) -- don't know if this is actually required
3. wait for a second or two
4. (maybe) see error

Expected behavior
a lack of crash

Version info:
commit=, build date=, build source=nix, version=0.50.0, os=linux, arch=amd64, git version=2.49.0
git version 2.49.0

Additional context
despite the backtrace seemingly talking about background fetches, this problem only occurs when I'm in the interactive rebase mode
Note: please try updating to the latest version or manually building the latest master to see if the issue still occurs.

[stefanhaller]

This is very strange, from the stack trace it looks like you have no branches, but I thought there must always be at least one (for the current head).

Could you post a screenshot of your lazygit window when you are in that interactive rebase (before it crashes)?

Anyway, here's a fix that should probably help, even if I can't explain why: #4565.

[ada4a]

apparently Lazygit thinks that I have no branches... but I of course do

[ada4a]

Yeah it seems like the branches disappear the moment I go into the interactive rebase mode. And how fast the crash happens depends purely on when the next round of the automatic fetch comes. (EDIT: no, if I just enter the interactive rebase mode and don't do anything, the refreshes don't yet cause the error. It's only after I move at least one commit that the next fetch kills Lazygit)

[stefanhaller]

Do the branches disappear the moment you enter the interactive rebase, or only once you move a commit?

[ada4a]

the moment I enter it

[stefanhaller]

In that case I'm pretty sure you will always get the crash once the next background fetch comes along, no matter whether you moved a commit first or not.

I suppose you should also be able to trigger the crash by going to the Files panel and pressing f (after entering the rebase).

[ada4a]

In that case I'm pretty sure you will always get the crash once the next background fetch comes along, no matter whether you moved a commit first or not.

Yes, you're right. It's just that I tried running lazygit --logs, and confused refreshes for fetches.

I suppose you should also be able to trigger the crash by going to the Files panel and pressing f (after entering the rebase).

Looks like that's right as well. It seems to run some Git commands, so the crash comes after 1-2 seconds. Here are the logs right after I press f -- hopefully, they can help a bit:

May 16 16:50:30 |INFO| git fetch --all --no-write-fetch-head (3.173551839s) 
May 16 16:50:30 |INFO| refreshing the following scopes in sync mode: branches,commits,remotes,tags 
May 16 16:50:30 |DEBU| using cache for key core.commentChar 
May 16 16:50:30 |DEBU| RunCommand command="git tag --list -n --sort=-creatordate"
May 16 16:50:30 |DEBU| using cache for key core.commentChar 
May 16 16:50:30 |INFO| git for-each-ref --sort=refname --format=%(refname:short) refs/remotes (4.331901ms) 
May 16 16:50:30 |INFO| postRefreshUpdate for remotes took 80.66µs 
May 16 16:50:30 |INFO| postRefreshUpdate for remoteBranches took 10.7µs 
May 16 16:50:30 |INFO| refreshed remotes in 5.149831ms 
May 16 16:50:30 |INFO| git tag --list -n --sort=-creatordate (5.091051ms) 
May 16 16:50:30 |INFO| postRefreshUpdate for tags took 13.045µs 
May 16 16:50:30 |INFO| refreshed tags in 5.542625ms 
May 16 16:50:30 |INFO| git -c log.showSignature=false show --no-patch --oneline --abbrev=20 --pretty=format:%H%x00%at%x00%aN%x00%ae%x00%D%x00%P%x00%m%x00%s 53e7cf1754f385acbcf058633a3dac229181e3d2 a10e6ff1c0c69678389851497ec395291cd3b695 fb7d96c09b668580f6355827967802c702c1bb34 17fd502cf5f9e63da7b5f70a873d69933ceec508 f0d211c21c4ef576978010c7fb2b5700756ffac0 6b3834053550895390b84593a8150f636e564956 349e2c66e17a9582dd64bc2241d68009f93e5c93 99a583c7c132f1771d516c9c6cd6df296351ca3c 0199e28b9826cb61ebc53757e320752c931ee12d (5.315671ms) 
May 16 16:50:30 |DEBU| RunCommand command="git merge-base HEAD refs/remotes/origin/main"
May 16 16:50:30 |DEBU| RunCommand command="git merge-base test-utils-game test-utils-game@{u}"
May 16 16:50:30 |INFO| git merge-base HEAD refs/remotes/origin/main (3.474117ms) 
May 16 16:50:30 |ERRO| fatal: no upstream configured for branch 'test-utils-game'
 command="git merge-base test-utils-game test-utils-game@{u}"
May 16 16:50:30 |INFO| git merge-base test-utils-game test-utils-game@{u} (6.500146ms) 
May 16 16:50:30 |INFO| git log HEAD --topo-order --oneline --pretty=format:%H%x00%at%x00%aN%x00%ae%x00%D%x00%P%x00%m%x00%s --abbrev=40 -300 --no-show-signature -- (10.6892ms) 
May 16 16:50:30 |DEBU| using cache for key rebase.updateRefs 
May 16 16:50:30 |INFO| postRefreshUpdate for commits took 432.078µs 
May 16 16:50:30 |INFO| refreshed commits and commit files in 17.515947ms 
May 16 16:50:30 |DEBU| using cache for key rebase.updateRefs 
May 16 16:50:30 |INFO| git -c log.showSignature=false log -g --abbrev=40 --format=%h%x00%ct%x00%gs%x00%P (21.875701ms) 
May 16 16:50:30 |INFO| postRefreshUpdate for reflogCommits took 101.741792ms 
May 16 16:50:30 |DEBU| RunCommand command="git for-each-ref --sort=-committerdate --format=%(HEAD)%00%(refname:short)%00%(upstream:short)%00%(upstream:track)%00%(push:track)%00%(subject)%00%(objectname)%00%(committerdate:unix) refs/heads"
May 16 16:50:30 |INFO| git for-each-ref --sort=-committerdate --format=%(HEAD)%00%(refname:short)%00%(upstream:short)%00%(upstream:track)%00%(push:track)%00%(subject)%00%(objectname)%00%(committerdate:unix) refs/heads (8.665085ms) 
May 16 16:50:30 |DEBU| RunCommand command="git symbolic-ref --short HEAD"
May 16 16:50:30 |ERRO| fatal: ref HEAD is not a symbolic ref
 
May 16 16:50:30 |INFO| git symbolic-ref --short HEAD (4.673871ms) 
May 16 16:50:30 |DEBU| RunCommand command="git branch --points-at=HEAD --format=%(HEAD)%00%(objectname)%00%(refname)"
May 16 16:50:30 |ERRO| fatal: unknown field name: commiterdate
 
May 16 16:50:30 |INFO| git branch --points-at=HEAD --format=%(HEAD)%00%(objectname)%00%(refname) (4.518169ms) 
May 16 16:50:30 |ERRO| fatal: unknown field name: commiterdate
 
May 16 16:50:30 |INFO| postRefreshUpdate for localBranches took 15.559µs 
May 16 16:50:30 |DEBU| using cache for key rebase.updateRefs 
May 16 16:50:30 |INFO| refreshed reflog and branches in 145.038394ms 
May 16 16:50:30 |INFO| Refresh took 145.139353ms 

(ignore the commiterdate thing -- it's an error in my Git config)

[stefanhaller]

What we need to find out is why lazygit suddenly starts to think you have no branches, and the logs don't show that.

Can you debug go code? I have this suspicion that this line might be the culprit (just a wild guess though). Would be cool if you could set a breakpoint there and see if you hit it when you are in the interactive rebase.

But I'm also curios about that commiterdate thing, can you tell us more about that?

[ada4a]

The list also empties out when I go to the Commits panel, and checkout a commit as detached HEAD (but not when the selected commit is a tip of an existing branch and I checkout it as such)

[ada4a]

What we need to find out is why lazygit suddenly starts to think you have no branches, and the logs don't show that.

Can you debug go code? I have this suspicion that this line might be the culprit (just a wild guess though). Would be cool if you could set a breakpoint there and see if you hit it when you are in the interactive rebase.

I don't have any experience with Go, but it can't be that hard. I'll see what I can do

But I'm also curios about that commiterdate thing, can you tell us more about that?

Oh it's just from a "cool things to put into your Git config"-kind of a video I saw some time ago. I guess I just misspelled the option...
Wait, I think it might actually be culprit here 😅 The place I have it in is branch.sort=-commiterdate. I think what's happening is that Lazygit tries to run git branch, but Git returns doesn't anything, because of this wrong config value. Therefore Lazygit concludes that we don't have any branches, and eventually dies.

Fwiw, that would still be a regression on Lazygit's side -- I've had this wrong thing set for a while now, but Lazygit only started reacting this way since the last update

[ada4a]

yeah, that seems to be it -- when I run git branch (which I don't really run without arguments all that often I guess), I get the following output:

fatal: unknown field name: commiterdate

how embarrassing 😅

[stefanhaller]

This could explain it, but I'm still confused why you only get the empty branches panel when inside an interactive rebase, and not always.

Fwiw, that would still be a regression on Lazygit's side -- I've had this wrong thing set for a while now, but Lazygit only started reacting this way since the last update

It's just the crash that is the regression. The branches panel going empty must have happened in earlier versions too.

BTW, instead of (or in addition to) changing the branch sort order globally in git, I recommend to change it in lazygit, by pressing s in the branches panel. Lazygit then has some optimizations that make it work a little better, because it no longer needs to maintain the recency data that it would otherwise sort by.

[ada4a]

it's not even an incorrect value! just a typo... it should've been committerdate, with 2 t's, see the second paragraph after this