ssh-agent: remove all keys upon SIGUSR1..
With the advent of per-user temporary directories it became hard for an administrator to remove all keys from all running ssh-agent instances; what formerly could be done like so
if command -v ssh-add >/dev/null 2>&1; then for a in /tmp/ssh-/agent.; do [ -e "$a" ] || continue act "SSH_AUTH_SOCK="$a" ssh-add -D </dev/null >/dev/null 2>&1 &" inc done fi
has become a major undertaking, especially with even more containerization. Being able to remove all keys from all agents with a single command seems so desirable that it is available in other agents in the software world.
Ah. Merde! Sorry, i thought "mergeable" meant the pull request!
..and one more sorry, this was not meant to be closed. I might become a github expert.
This looks good to me and the better signal handling could allow us to make cleanup_handler() a regular call in future (i.e. no longer calling it in signal context).
Would appreciate @daztucker taking a look at this too
(It is all taken from sshd.c of course. Only to mention it. I .. do not implement this cleanup like it is done in sshd.c in this context here now!?!) Thank you for looking into this.
But surely not because of this changeset, no? Never used CI as it is here; but there is no log at all for the failure, not even a build one?
(i do not understand this CI-fuzz github thing which fails in one out of dozens of things. this should have nothing to do with the little patch, i would think.)
Interestingly "the ppoll part has just recently been committed" almost as stated here. Please let me allow one sentence: what a crappy attitude to leave that truly desired piece of code laying around for two years. git allows selective staging via "git add -p", and you would have had it 22 months ago. Anyhow, this little pull request is now solely the SIGUSR1 handling, and all tests ran.
P.S.: "you would have had it 22 months ago": within less than 30 seconds.