Closed 20kdc closed 3 years ago
Should be good as of 017735f, sorry I have brain rot.
Let me know
Well, a change almost exactly equivalent in text (slight spacing difference) to that commit was already being tested, and the zombie problem is apparently now solved in all the places where init is supposed to be doing the zombie-reaping, so yeah, should be fine also, thank you
See #14 Zombies remain very unreaped, which appears to have caused trouble for a friend of mine using FVWM. I have also confirmed this by constructing an environment using the following (very terrible) script and creating zombies by nesting sh processes and killing them.
This was tested under commit 6ba337adc4aec900e6e2b4df287676378fe5d746.
To create a zombie:
Then from outside the namespace, check top:
The entry with a Z here is the zombie. Alternatively,
ps -e
will show a process as13595 pts/2 00:00:00 sh <defunct>
. (This is from a different run, so a different PID.)The zombie process will be cleaned up by the outer init when hummingbird terminates.
The zombies can be reaped by adding
signal(SIGCHLD, handler);
past the other twosignal
calls inhummingbird.c
.