"error: getting sandbox mount namespace: No such file or directory" on some commands

[NobbZ]

Describe the bug

I have not yet found out why exactly this happens and under which circumstances.

But nixos-rebuild as well as home-manager switch fail eventually with either staling or the error mentioned in the subject.

With nixos-rebuild switch this happened early and without much other messages, while with HM it happens during activation:

$ # Here I tried to get my current config but with a generally known to work nix, to rule
$ # out problems with the demon, I eventually rolled back to a good configuration.
$ sudo nixos-rebuild switch --flake github:nobbz/nixos-config --override-input nix github:nixos/nix/2.14.0
warning: not writing modified lock file of flake 'github:nobbz/nixos-config':
• Updated input 'nix':
    'github:nixos/nix/b5b6055bbf00fa809633e88dab1e8eaec3ae9f8d' (2023-03-18)
  → 'github:nixos/nix/4b156ad944bd776eeeba4b7394781ac02c68d7a4' (2023-02-28)
• Removed input 'nix/flake-compat'
[0/1 built, 1.2 MiB DL]
error: getting sandbox mount namespace: No such file or directory
$ home-manager switch --flake github:nobbz/nixos-config                                                                                                                                                                                                                                                 
Starting Home Manager activation                                                                                                                                                                                                                                              
Activating checkFilesChanged                                                                                                                                                                                                                                                  
Activating checkLinkTargets                                                                                                                                                                                                                                                   
Activating writeBoundary                                                                                                                                                                                                                                                      
Activating installPackages                                                                                                                                                                                                                                                    
replacing old 'home-manager-path'                                                                                                                                                                                                                                             
installing 'home-manager-path'                                                                                                                                                                                                                                                
error: getting sandbox mount namespace: No such file or directory                                                                                                                                                                                                                                              
                                                                                                                                                                                                                                                                              
Oops, Nix failed to install your new Home Manager profile!                                                                                                                                                                                                                    
                                                                                                                                                                                                                                                                              
Perhaps there is a conflict with a package that was installed using                                                                                                                                                                                                           
"nix-env -i"? Try running                                                                                                                                                                                                                                                     
                                                                                                                                                                                                                                                                              
    nix-env -q                                                                                                                                                                                                                                                                
                                                                                                                                                                                                                                                                              
and if there is a conflicting package you can remove it with                                                                                                                                                                                                                  
                                                                                                                                                                                                                                                                              
    nix-env -e {package name}                                                                                                                                                                                                                                                 
                                                                                                                                                                                                                                                                              
Then try activating your Home Manager configuration again.

Steps To Reproduce

1. Use nix from b5b6055
2. try to switch the configuration as above. I had the error randomly appearing on build and shell as well
3. see error

Expected behavior

No error and no staling build

nix-env --version output

$ # bad
$ nix --version
nix (Nix) 2.15.0pre20230318_b5b6055
$ # good
$ nix --version
nix (Nix) 2.15.0pre20230315_eb56cb7

Additional context

Not yet bisected. Especially as I am not quite sure if this is a single problem with multiple symptoms or 2 or more individual problems (stale and error)

edit: (A forced build seems to trigger at least one instance of the error and I start bisecting that now under the assumption that 30 second "stale" are the same as the error message from subject)

Priorities

Add 👍 to issues you find important.

[NobbZ]

Under the assumption that "stale" and "error" are the same, I find c536e00 as the culprit. But except for the initial BAD actually all were stale.

git bisect log

git bisect start
# status: waiting for both good and bad commits
# bad: [b5b6055bbf00fa809633e88dab1e8eaec3ae9f8d] Merge pull request #8064 from knedlsepp/flake-in-submodule
git bisect bad b5b6055bbf00fa809633e88dab1e8eaec3ae9f8d
# status: waiting for good commit(s), bad commit known
# good: [eb56cb7cc7fa9fb58e2d9dc0322375da23b0eb65] Merge pull request #7750 from obsidiansystems/no-args-prepare
git bisect good eb56cb7cc7fa9fb58e2d9dc0322375da23b0eb65
# bad: [7f46ebcf90432a54f5fdec1931d87e5130e68190] Merge pull request #8049 from edolstra/unexpected-eof
git bisect bad 7f46ebcf90432a54f5fdec1931d87e5130e68190
# bad: [6029c763c2c5998dc3265152425c8ff0ce01b1a0] Change builderOut from Pipe to AutoCloseFD
git bisect bad 6029c763c2c5998dc3265152425c8ff0ce01b1a0
# bad: [c536e00c9deeac58bc4b3299dbc702604c32adbe] Move pseudoterminal slave open to child
git bisect bad c536e00c9deeac58bc4b3299dbc702604c32adbe
# good: [19326ac2979f0d989835105a5d816a943a6bc7f2] Simplify commonChildInit()
git bisect good 19326ac2979f0d989835105a5d816a943a6bc7f2
# first bad commit: [c536e00c9deeac58bc4b3299dbc702604c32adbe] Move pseudoterminal slave open to child

Individual commits have been tested using nix build . && sudo ./result/bin/nix build nixpkgs#hello --rebuild.

I will also try a bisection now where I git bisect skip the "stale" ones.

[NobbZ]

Skipping the stale ones:

There are only 'skip'ped commits left to test.
The first bad commit could be any of:
c536e00
cb28e4f
We cannot bisect more!

git bisect log

git bisect start
# status: waiting for both good and bad commits
# bad: [b5b6055bbf00fa809633e88dab1e8eaec3ae9f8d] Merge pull request #8064 from knedlsepp/flake-in-submodule
git bisect bad b5b6055bbf00fa809633e88dab1e8eaec3ae9f8d
# status: waiting for good commit(s), bad commit known
# good: [eb56cb7cc7fa9fb58e2d9dc0322375da23b0eb65] Merge pull request #7750 from obsidiansystems/no-args-prepare
git bisect good eb56cb7cc7fa9fb58e2d9dc0322375da23b0eb65
# skip: [7f46ebcf90432a54f5fdec1931d87e5130e68190] Merge pull request #8049 from edolstra/unexpected-eof
git bisect skip 7f46ebcf90432a54f5fdec1931d87e5130e68190
# skip: [bfb9eb87fe98f96969188df9df866e15800bd55b] Cleanup      test skipping
git bisect skip bfb9eb87fe98f96969188df9df866e15800bd55b
# skip: [c536e00c9deeac58bc4b3299dbc702604c32adbe] Move pseudoterminal slave open to child
git bisect skip c536e00c9deeac58bc4b3299dbc702604c32adbe
# skip: [5eb8bfd0f17f950ec181d59fb9fbb2330f778935] Remove a variable in LocalDerivationGoal
git bisect skip 5eb8bfd0f17f950ec181d59fb9fbb2330f778935
# skip: [f9c24d67b96f7d2cc4293ac8c3fbed70eb8328a8] Add a test with flake.nix in a submodule
git bisect skip f9c24d67b96f7d2cc4293ac8c3fbed70eb8328a8
# skip: [acd707acca8acbc276af5181c7724c70a6a69ad5] Merge pull request #8057 from obsidiansystems/slight-test-cleanup
git bisect skip acd707acca8acbc276af5181c7724c70a6a69ad5
# good: [581e11cd552152c27540de23e619aa61900ef9f4] Merge pull request #8053 from DeterminateSystems/fix-structuredattrs-sandbox
git bisect good 581e11cd552152c27540de23e619aa61900ef9f4
# good: [19326ac2979f0d989835105a5d816a943a6bc7f2] Simplify commonChildInit()
git bisect good 19326ac2979f0d989835105a5d816a943a6bc7f2
# skip: [6029c763c2c5998dc3265152425c8ff0ce01b1a0] Change builderOut from Pipe to AutoCloseFD
git bisect skip 6029c763c2c5998dc3265152425c8ff0ce01b1a0
# bad: [cb28e4fe2a552000272793e248ea3f6a69d625d6] Remove "unexpected EOF" retry hack
git bisect bad cb28e4fe2a552000272793e248ea3f6a69d625d6
# only skipped commits left to test
# possible first bad commit: [cb28e4fe2a552000272793e248ea3f6a69d625d6] Remove "unexpected EOF" retry hack
# possible first bad commit: [c536e00c9deeac58bc4b3299dbc702604c32adbe] Move pseudoterminal slave open to child

And as cb28e4f only removes code from the testrunner I am absolutely not sure how it could affect the result of a nix build … --rebuild…

As I have not enough knowledge about C++ or any of the affected things nor any of the related issues I will leave drawing of conclusions from these 2 bisections to anyone who gets assigned to solving this issue.

[cole-h]

@NobbZ Eelco opened #8077 for this -- could you see if it fixes your issue?

[cole-h]

I think this can be closed now that #8077 has merged :)

[NobbZ]

Definitely.