mate-screenshot orphans subprocess tree after closing

[thanatos]

Steps to Reproduce

1. Run mate-screenshot --window. That will create a process tree something like the following:

   /bin/zsh
   └─ mate-screenshot --window
      └─ mate-screenshot --window
         └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev --ro-bind-try
            └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev --ro-bind-
               └─ /usr/lib/glycin-loaders/2+/glycin-image-rs --dbus-fd 28
   

2. Close the window that appears / close/quit mate-screenshot

3. The subprocess tree still exists:

   /sbin/init
   ⁞
   ├─ mate-screenshot --window
   │  └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr 
   │     └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /u
   │        └─ /usr/lib/glycin-loaders/2+/glycin-image-rs --dbus-fd 28
   

   While the main process appears to have exited, the subprocess it spawned has been reparented to init, and appears to run indefinitely, leaking 4 processes and the associated RAM & other resources.

Expected Behavior

mate-screenshot does not orphan/leak child processes.

Actual Behavior

mate-screenshot orphans a child process.

Error Output

Version

mate-screenshot 1.28.0

Distribution

Arch Linux

Additional Context

No response

Reproducibility

Every time

Checklist

• I have searched existing issues to avoid duplicates
• I have provided all the requested information above

[vkareh]

This is not really a mate-screenshot bug, but a bug in Bubblewrap (the sandboxing tool used by Glycin, GTK's image loader)

• https://gitlab.gnome.org/GNOME/glycin/-/work_items/312
• child bwrap process is left as zombie containers/bubblewrap#697

You could try validating on your side by running the following command:

$ GLYCIN_DISABLE_SANDBOX=i-know-the-risks mate-screenshot --window

(see https://gitlab.gnome.org/GNOME/glycin/-/commit/3f44c7e943ca6b7c291a8eedddd30fcdb750571f for reference)

[thanatos]

Hmm. So, I didn't think this was necessarily a glycin bug because the most ancestral orphan is mate-screenshot, not bwrap etc. (But, it is a bunch of brwap & glycin processes, so it could be, too.)

The leaked process tree in my case are not zombies; they're all live processes:

3183584 me          20   0 1850M 19776  5300 S   0.0  0.1  0:00.00 ├─ mate-screenshot --window
3183586 me          20   0  2856  1992  1852 S   0.0  0.0  0:00.00 │  └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev --
3183587 me          20   0  2856  1352  1188 S   0.0  0.0  0:00.00 │     └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev
3183588 me          20   0 76448  4760  4420 S   0.0  0.0  0:00.00 │        └─ /usr/lib/glycin-loaders/2+/glycin-image-rs --dbus-fd 28
3184435 me          20   0 1850M 19632  5148 S   0.0  0.1  0:00.00 ├─ mate-screenshot --window
3184437 me          20   0  2856  2116  1976 S   0.0  0.0  0:00.00 │  └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev --
3184438 me          20   0  2856  1376  1216 S   0.0  0.0  0:00.00 │     └─ bwrap --unshare-all --die-with-parent --chdir / --ro-bind /usr /usr --dev /dev
3184439 me          20   0 76448  4728  4388 S   0.0  0.0  0:00.00 │        └─ /usr/lib/glycin-loaders/2+/glycin-image-rs --dbus-fd 28

The whole orphaned tree is still alive.

This,

GLYCIN_DISABLE_SANDBOX=i-know-the-risks mate-screenshot --window

Doesn't seem to have any effect.

[vkareh]

Wait, I just noticed that your report mentions 1.28.0
Can you check with 1.28.1? I think this is something we fixed last year with c74d86d

[thanatos]

Can you check with 1.28.1?

I'm on Arch, so the simplest way would be to do an update there. It seems like the relevant Arch package hasn't updated yet, though. If I have time, I'll see about doing it locally, from scratch.

(The is an open bug on the Arch side, at least, for the package being out of date w/ upstream.)