hyprland-wiki/pages/Crashes and Bugs/_index.md

138 lines
3.8 KiB
Markdown

---
weight: 10
title: Crashes and Bugs
---
## Getting the log
If you are in a TTY, and the Hyprland session that crashed was the last one you
launched, the log can be printed with
```sh
cat $XDG_RUNTIME_DIR/hypr/$(ls -t $XDG_RUNTIME_DIR/hypr/ | head -n 1)/hyprland.log
```
if you are in a Hyprland session, and you want the log of the last session, use
```sh
cat $XDG_RUNTIME_DIR/hypr/$(ls -t $XDG_RUNTIME_DIR/hypr/ | head -n 2 | tail -n 1)/hyprland.log
```
## Obtaining the Hyprland Crash Report
If you have `$XDG_CACHE_HOME` set, the crash report directory is
`$XDG_CACHE_HOME/hyprland`. If not, it's `$HOME/.cache/hyprland`.
Go to the crash report directory and you should find a file named
`hyprlandCrashReport[XXXX].txt` where `[XXXX]` is the PID of the process that
crashed.
Attach that file to your issue.
## Crashes at launch
Diagnose the issue by what is in the log:
- `sWLRBackend was NULL!` -> launch in the TTY and refer to the wlr logs in RED.
- `Monitor X has NO PREFERRED MODE, and an INVALID one was requested` -> your
monitor is bork.
- Other -> see the coredump. Use `coredumpctl`, find the latest one's PID and do
`coredumpctl info PID`.
- failing on a driver (e.g. `radeon`) -> try compiling with
`make legacyrenderer`, if that doesn't help, report an issue.
- failing on `wlr-xxx` -> try compiling with `make legacyrenderer`, if that
doesn't help, report an issue, and also refer to the TTY wlr logs in RED like
in the first point.
- failing on `Hyprland` -> report an issue.
## Crashes not at launch
Report an issue on GitHub or on the Discord server.
## Obtaining a debug stacktrace
Systemd-only.
Build hyprland in debug (`make debug`) and run. Get it to crash. Then, in a tty or terminal, do
`coredumpctl debug Hyprland`.
If gdb asks you for symbols, say `y`, if it asks about paging, say `c`.
Once you get to `(gdb)`, run `bt -full` and post the output.
## Bugs
First of all, **_READ THE [FAQ PAGE](../FAQ)_**
If your bug is not listed there, you can ask on the Discord server or open an
issue on GitHub.
## Bisecting an issue
"Bisecting" is finding the first _git_ commit that introduced a specific bug or
regression using binary search. This is done in `git` using the `git bisect` command.
First, clone the Hyprland repo if you haven't already:
```sh
git clone --recursive https://github.com/hyprwm/Hyprland
cd Hyprland
```
Start the bisect process:
```sh
git bisect start
```
Enter the first known good commit hash that did not contain the issue:
```sh
git bisect good [good commit]
```
Then, enter the known bad commit hash that does contain the issue. You can simply use HEAD:
```sh
git bisect bad HEAD
```
_git_ will now checkout a commit in the middle of the specified range.
Now, reset and build Hyprland:
```sh
git reset --hard --recurse-submodules
make all
```
...and run the built executable from the TTY `./build/Hyprland`.
Try to reproduce your issue. If you can't (i.e. the bug is not present), go back to the
Hyprland repo and run `git bisect good`. If you can reproduce it, run `git bisect bad`.
_git_ will then checkout another commit and continue the binary search.
If there's a build error, run `git bisect skip`.
Reset, build and install Hyprland again and repeat this step until _git_ identifies the
commit that introduced the bug:
```
[commit hash] is the first bad commit
```
## Building the Wayland stack with ASan
If requested, this is the deepest level of memory issue debugging possible.
_Do this in the tty, with no Hyprland instances running._
Clone hyprland: `git clone --recursive https://github.com/hyprwm/Hyprland`
`make asan`
Reproduce your crash. Hyprland will exit back to the tty.
Now, in either `cwd`, `~` or `./build`, search for file(s) named
`asan.log.XXXXX` where XXXXX is a number.
Zip all of them up and attach to your issue.