bash complains wait: pid is not a child of this shell

[Scott Moser]

inside a container the attached script will often fail, with bash claiming:
 wait: pid 13785 is not a child of this shell

I've been able to reproduce this in clean vms (lxd init, then lxc launch) and on my desktop.
  xenial host (kernel 4.4.0-22-generic) and xenial container
  yakkety host (kernel 4.4.0-23-generic) and xenial container

Containers are created simply with:
   lxd launch ubuntu-daily:xenial x1

xenial lxd was 2.0.2-0ubuntu1~16.04.1 , yakkety was 2.0.1-0ubuntu1.

I've not been able to recreate at all outside of a container.

The test case basically does:
  # spawn NUM_CHILDREN (default 200) children in backgrounded subprocesses.
  # then 'wait' for each of them to return in spawned order and
  # check exit code. Exit failure if any 'wait' return non-zero
  # when 'pid XXX is not a child of this shell' is printed, 127 is found.

To run it in a loop until it fails:
  # run this in a loop until it fails with:
  # export NUM_RUNS=200 ; while :; do rm -f out*; /tmp/go.sh || break; done

Note, that increasing NUM_RUNS seems to increase likelyhood of failure.

I'm not sure if this is a bash bug or container bug or kernel bug.
I've filed against lxd, feel free to move it where ever.

ProblemType: Bug
DistroRelease: Ubuntu 16.10
Package: lxc2 (not installed)
ProcVersionSignature: User Name 4.4.0-23.41-generic 4.4.10
Uname: Linux 4.4.0-23-generic x86_64
NonfreeKernelModules: zfs zunicode zcommon znvpair zavl
ApportVersion: 2.20.1-0ubuntu4
Architecture: amd64
Date: Tue Jun 7 13:14:52 2016
Ec2AMI: ami-000001a7
Ec2AMIManifest: FIXME
Ec2AvailabilityZone: nova
Ec2InstanceType: m1.small
Ec2Kernel: unavailable
Ec2Ramdisk: unavailable
ProcEnviron:
 TERM=xterm-256color
 PATH=(custom, no user)
 XDG_RUNTIME_DIR=<set>
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: lxd
UpgradeStatus: No upgrade log present (probably fresh install)

[Scott Moser (smoser)]

I put https://gist.github.com/smoser/68f4f1d379fc55b64d2d2cd33922fe7b together.
it incorporates the looping so you can just run:
 ./bug-1590001 loop

or
 ./bug-1590001 spawnkids sleep-rand
or
 ./bug-1590001 spawnkids sleep-short

[Stéphane Graber (stgraber)]

So trying to remember from IRC, I think we came to the conclusion that this isn't a LXD bug but most likely some weird bash bug that's weirdly enough, much easier to hit in a container.

Do you want this re-assigned to the bash package instead or just closed entirely?