Warp exhausts file watchers

[zmrocze]

Pre-submit Checks

• I have searched Warp bugs and there are no duplicates
• I have searched Warp known issues page and my issue is not there
• I have an issue with AI and have included the debugging ID (Optional, but helps expedite the AI quality fix). Debugging ID intructions
• I have technical issue and have included the logs (optional, but helps expedite the bug fix). Log intructions

Describe the bug

Hey,

with single indexed directory it seems like Warp exhausts all available file watchers using 523_603 of them. The issue is not observed with warp, but later i.e. when opening vscode or windsurf. It depends on warp indexing being turned on or off.

I have both .gitignore and .warpindexignore files. Excluding files mentioned there - there is only 16 files in the directory. Including all files there are ~15000 files.
Im having some trouble resolving watched file names, hopefully ill get back to it tomorrow.

How is that possible?
I'd also like more clarity on what is being indexed (any paths found in $PATH maybe?) and the ability to specify .warpindexingignore in settings (so globally).

Thanks!!!

To reproduce

1. open warp and index a directory
2. open vscode and observe "enospc" error

Expected behavior

No response

Screenshots, videos, and logs

No response

Operating system (OS)

Linux

Operating system and version

nixos 25.04

Shell Version

No response

Current Warp version

Warp version: Some("v0.2025.08.20.08.11.stable_03")

Regression

Yes, this bug started recently or with an X Warp version

Recent working Warp date

No response

Additional context

No response

Does this block you from using Warp daily?

Yes, this issue prevents me from using Warp daily.

Is this an issue only in Warp?

Yes, I confirmed that this only happens in Warp, not other terminals.

Warp Internal (ignore): linear-label:b9d78064-c89e-4973-b153-5178a31ee54e

None

[advait-m]

Hmm, weird, thanks for reporting - could you provide Warp logs (https://docs.warp.dev/support-and-billing/sending-us-feedback#gathering-warp-logs) - either here or email (I'm advait at warp.dev)

[zmrocze]

Hey, sent logs. and checked what are these watched files

The watched files are:

• few expected from my indexed directory
• 174563 files from /nix/store. The /proc/sys/fs/inotify/max_user_watches limit of 524288 is close to 3 times this and these files are watched by 3 warp threads.
• additionally few unexpected from a directory that was once indexed but is no more

/nix/store

I'm on nixos system. It installs all programs into a global /nix/store. Development dependencies are provided through env variables ($PATH, $PYTHON_PATH, etc) - I'm using devenv.sh with direnv.

[zmrocze]

Guys, what about this? You can't just index my whole /nix/store -_-

/nix/store is clearly outside of the working directory. Nor is it symlinked anywhere among the non .gitignore'd files.
So do you index files referenced in environment variables, like PATH? Does that proceed recursively?
That's my guess on what happens. Can you verify this is not? Then we're in woods. But you cannot index my whole /nix/store. This exhausts 500k filewatchers.

@advait-m could i ping you?

[mbalc]

can confirm, in my case with the defaults of:

  max_user_watches     65,536

Warp takes up around 2/3 of the total limit:

       Pid Uid        App                            Watches  Instances
   1296318 1000       warp                            41,092          5

do you know of some workaround for this, is there some feature I could disable to free these watches for now, aside from quitting Warp?

sending you my logs too via email

[mbalc]

I'm checking now what files are being watched on my Debian system, scrolling through - and in my case, I see watches on individual files/directories of stuff like:

• lots of files on the repo I work on, biggest offenders including:
  • node_modules (!)
  • .yarn/index
  • our dist/ folder with the project build
  • .terraform/modules

[zmrocze]

note that in my case the watched files are even completely outside of the workspace (in /nix/store)

[AlfreddGco]

This happens to me too in Debian 13 system. Kind of bothering since restarting warp doesn't always work

[EricTheMagician]

I'm also having this problem with nixos.
I think I have the same problem with fedora (but I use nix-shell there).

[benjamin-cooper-ai]

I have this problem without nixos, normal Ubuntu 22 system. I even have AI stuff off so I'm not sure what the files are being watched for.

[AlfreddGco]

My current solution is running npm/node processes on a different terminal and use warp for AI and nvim. This has been working for me 90% of the time. It's fine if I keep the context separation as: "x" terminal runs everything, and warp I do everything else. But it would still be great if Warp solved this issue

[mbalc]

did someone take a deeper look at this yet? i moved from gnome to kde and everyday i start my workstation, i get prompted to and I do bump the inotify watcher limit little by little

is there some workaround or something like, idk, .warpignore-like setting?

[bcheidemann]

I am also experiencing this issue on a normal Ubuntu setup (default limits for watchers).

I have codebase index off and all AI features off, and yet Warp registers nearly 35k watchers whenever I cd into my monorepo. Worse still, as I cd around, it doesn't seem to drop old watchers, so the whole thing snowballs.

I really love warp, and it's cool that you are trying to build new features into terminals. But in my opinion, registering tens of thousands of watchers is not a good default experience, especially when there seems to be no off-switch.

[benjamin-cooper-ai]

It's not even related to usage, it just monotonically goes up from start. I'm done with Warp until this is fixed.

[stabai]

Yeah, it seems completely unusable to me. Sad because I really like the app a lot and have gotten used to it over the years on Mac machines.