fish appears to hang on first startup

[mgedmin]

I've been meaning to try out fish on several occasions (years apart), but I always stumble upon this issue.

On my system (various versions of Ubuntu, currently 19.10) I do

• sudo apt install fish
• fish

and what happens next:

• no output
• no prompt
• laptop's fan spins up
• nothing else happens for several minutes
• I panic and kill fish that is eating all the CPU

On second run fish starts instantly.

I'm not sure how to reset the state so this happens again.

fish version: 3.0.2
OS: Ubuntu 19.10
$TERM: xterm-256color
Terminal emulator: gnome-terminal 3.34.0 (VTE 0.58.2)

I cannot reproduce this with sh -c 'env HOME=$(mktemp -d) fish'! But if I try to close the shell I spawned that way, I get a warning about active jobs, which might be a Clue.

Welcome to fish, the friendly interactive shell
mg@blynas /h/m/s/z/BTrees> 
There are still jobs active:

   PID  Command
 25573  $py $update_args >/dev/null 2>&1 &

A second attempt to exit will terminate them.
Use 'disown PID' to remove jobs from the list without terminating them.
mg@blynas /h/m/s/z/BTrees> 

(this also leaves my terminal in a noecho mode, which is annoying but unrelated)

It seems that something in my home directory (its sheer size? some config file somewhere?) is causing this, but I don't know what.

[mgedmin]

I've discovered a way of reproducing this:

• rm -rf ~/.local/share/fish/
• fish

Now fish is sitting there doing... something... with no output and no input, eating 100% of one CPU core. pstree shows it has no subproesses. strace shows it doing weird lstat calls on files in my home directory

...
getdents64(76, /* 14 entries */, 32768) = 400
brk(0x5592f9003000)                     = 0x5592f9003000
getdents64(76, /* 0 entries */, 32768)  = 0
lseek(76, 0, SEEK_SET)                  = 0
getdents64(76, /* 14 entries */, 32768) = 400
stat("PlayOnLinux's virtual drives/Subnautica/dosdevices/z:/sys/kernel/software_nodes/node0/0-0015/subsystem/devices/i2c-3/device/i2c-1/i2c-dev/i2c-1/subsystem/i2c-0/device/firmware_node/physical_node/iTCO_wdt/subsystem/devices/PNP0C14:00/firmware_node/subsystem/devices/device:2d/physical_node1/2-2.3/port/peer/device/1-2.3:1.0/driver/1-2:1.0/1-2-port1/peer/device/2-2.1:2.1/driver/2-2.1:2.0/net/enx482ae33b002a/subsystem/wlp0s20f3/phy80211/rfkill2/subsystem/rfkill0/device/leds/platform::mute/subsystem/phy0-led/device/subsystem/devices/0000:3a:00.0/usb4/subsystem/drivers/usb/1-2.3.3/1-2.3.3:1.0/1-2.3.3-port2/device/1-2.3.3.2:1.3/0003:17EF:A396.0019/input/input39/uevent", {st_mode=S_IFREG|0644, st_size=4096, ...}) = 0
stat("PlayOnLinux's virtual drives/Subnautica/dosdevices/z:/sys/kernel/software_nodes/node0/0-0015/subsystem/devices/i2c-3/device/i2c-1/i2c-dev/i2c-1/subsystem/i2c-0/device/firmware_node/physical_node/iTCO_wdt/subsystem/devices/PNP0C14:00/firmware_node/subsystem/devices/device:2d/physical_node1/2-2.3/port/peer/device/1-2.3:1.0/driver/1-2:1.0/1-2-port1/peer/device/2-2.1:2.1/driver/2-2.1:2.0/net/enx482ae33b002a/subsystem/wlp0s20f3/phy80211/rfkill2/subsystem/rfkill0/device/leds/platform::mute/subsystem/phy0-led/device/subsystem/devices/0000:3a:00.0/usb4/subsystem/drivers/usb/1-2.3.3/1-2.3.3:1.0/1-2.3.3-port2/device/1-2.3.3.2:1.3/0003:17EF:A396.0019/input/input39/capabilities", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/home", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
lstat("/home/mg", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
lstat("/home/mg/PlayOnLinux's virtual drives", {st_mode=S_IFLNK|0777, st_size=34, ...}) = 0
readlink("/home/mg/PlayOnLinux's virtual drives", "/home/mg/.PlayOnLinux//wineprefi"..., 4095) = 34
lstat("/home", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
lstat("/home/mg", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
lstat("/home/mg/.PlayOnLinux", {st_mode=S_IFDIR|0775, st_size=4096, ...}) = 0
lstat("/home/mg/.PlayOnLinux/wineprefix", {st_mode=S_IFDIR|0775, st_size=4096, ...}) = 0
lstat("/home/mg/.PlayOnLinux/wineprefix/Subnautica", {st_mode=S_IFDIR|0775, st_size=4096, ...}) = 0
lstat("/home/mg/.PlayOnLinux/wineprefix/Subnautica/dosdevices", {st_mode=S_IFDIR|0775, st_size=4096, ...}) = 0
lstat("/home/mg/.PlayOnLinux/wineprefix/Subnautica/dosdevices/z:", {st_mode=S_IFLNK|0777, st_size=1, ...}) = 0
readlink("/home/mg/.PlayOnLinux/wineprefix/Subnautica/dosdevices/z:", "/", 4095) = 1
lstat("/sys", {st_mode=S_IFDIR|0555, st_size=0, ...}) = 0
lstat("/sys/kernel", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/kernel/software_nodes", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/kernel/software_nodes/node0", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/kernel/software_nodes/node0/0-0015", {st_mode=S_IFLNK|0777, st_size=0, ...}) = 0
readlink("/sys/kernel/software_nodes/node0/0-0015", "../../../devices/pci0000:00/0000"..., 4095) = 53
lstat("/sys/devices", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/devices/pci0000:00", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/devices/pci0000:00/0000:00:1f.4", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/devices/pci0000:00/0000:00:1f.4/i2c-0", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/devices/pci0000:00/0000:00:1f.4/i2c-0/0-0015", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/devices/pci0000:00/0000:00:1f.4/i2c-0/0-0015/subsystem", {st_mode=S_IFLNK|0777, st_size=0, ...}) = 0
readlink("/sys/devices/pci0000:00/0000:00:1f.4/i2c-0/0-0015/subsystem", "../../../../../bus/i2c", 4095) = 22
lstat("/sys/bus", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/bus/i2c", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/bus/i2c/devices", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
lstat("/sys/bus/i2c/devices/i2c-3", {st_mode=S_IFLNK|0777, st_size=0, ...}) = 0
readlink("/sys/bus/i2c/devices/i2c-3", "../../../devices/pci0000:00/0000"..., 4095) = 46
...

I think it's getting confused when files have a : in the filename. on closer look that actually is a symlink to the root directory, I wonder if fish is getting into an infinite loop?

Note that ^C doesn't interrupt fish, I have to pkill it.

[mgedmin]

I tried rm -rf ~/.local/share/fish; strace -o /tmp/fish.trace fish and looked at the trace, and it looks like fish is trying to import my bash history? But it does that by opening and closing ~/.local/share/fish/fish_history once for every single line of history?

My ~/.bash_history has 25876 lines in it. fish managed to call open() on fish_history 13723 times in the first 16 seconds, before I killed it.

(My bashrc sets HISTSIZE=50000 and HISTFILESIZE=50000.)

[krobelus]

A workaround is to create an empty ~/.local/share/fish/fish_history so fish will not import bash history, but we should definitely make this more efficient

[krobelus]

@mgedmin This commit might solve it: krobelus@speedup-bash-history-import, I was able to import 20k lines of history instantly.

[krobelus]

Fixed in master now. Thanks for your very useful report!

You can remove your ~/.local/share/fish/fish_history and start a version of fish with the fix to import your bash history.

[mgedmin]

I can confirm that this fixes the entirety of the problem. Thank you!