summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Mark Wooding [Wed, 9 Aug 2017 02:59:04 +0000 (03:59 +0100)]
with-authinfo-kludge: Declare this an actual version.
It's been running with proper users and no complaints for over a year,
so I think I can safely say that it's working well enough.
Mark Wooding [Wed, 9 Aug 2017 02:58:16 +0000 (03:58 +0100)]
with-authinfo-kludge: Fix a poorly formatted comment.
Mark Wooding [Wed, 9 Aug 2017 02:57:07 +0000 (03:57 +0100)]
with-authinfo-kludge.1: Give the real reason for the `+' marker.
Mark Wooding [Wed, 9 Aug 2017 02:41:29 +0000 (03:41 +0100)]
with-authinfo-kludge.1: Fix garbled introductory text.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Actually write pidfiles like the manual says.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Change how ssh(1) signals that it's started.
* In `server_listen', arrange for the remote shell echo something to
stdout. Also, use `read' rather than cat(1), which saves a process
on the remote system. And, finally, include a comment so that the
ps(1) output is approximately useful.
* In `wait_for_ssh', go into a loop waiting for the echoed stuff from
the remote servers to arrive (or for them to report early EOF).
Thanks to Ian Jackson for suggesting this approach: it's much better
than the previous one.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Make $CLIENTKID be defined.
Otherwise Perl tells me off.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Make the listening sockets be non-blocking.
There's the traditional race if you don't, where a client comes along,
connects, and dies before you get around to accepting.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge (accept_loop): Might as well use the loop name.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge (accept_loop): Fix `select' error handling.
We shouldn't go through the file descriptors if select(2) failed: the
table is likely to be bobbins.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Remove done things from to-do list.
Mark Wooding [Mon, 25 Apr 2016 01:14:42 +0000 (02:14 +0100)]
with-authinfo-kludge: Trim default port from `authinfo-kludge' argument.
Only my specially hacked version can actually cope with this.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge: Add in some missing sigils.
Perl 5.20 helpfully warned about these where 5.14 didn't bother.
Mark Wooding [Sun, 24 Apr 2016 22:31:49 +0000 (23:31 +0100)]
with-authinfo-kludge, with-authinfo-kludge.1: Implementation!
It mostly seems to work, but don't lean on it too heavily yet, please!
I've made some very sketchy notes in the manpage about things I've
changed my mind about while writing the program. These will be turned
into actual text later.
Mark Wooding [Sun, 24 Apr 2016 22:30:15 +0000 (23:30 +0100)]
with-authinfo-kludge.1: Explain that we don't configure the client.
Mark Wooding [Sun, 24 Apr 2016 22:30:30 +0000 (23:30 +0100)]
with-authinfo-kludge.1: Clarifications about finding the runtime dir.
Mark Wooding [Sun, 24 Apr 2016 22:29:31 +0000 (23:29 +0100)]
with-authinfo-kludge.1: Zap some trailing whitespace.
It's like brambles. If you let it, it'll take over.
Mark Wooding [Sat, 23 Apr 2016 18:57:17 +0000 (19:57 +0100)]
with-authinfo-kludge.1: Remove bug paragraph about multiple runtime dirs.
Even I can't imagine why anyone would want more than `-d' provides.
A result of feedback from Ian Jackson.
Mark Wooding [Sat, 23 Apr 2016 18:56:31 +0000 (19:56 +0100)]
with-authinfo-kludge.1: Don't bother digging into the password database.
A result of feedback from Ian Jackson.
Mark Wooding [Sat, 23 Apr 2016 16:50:15 +0000 (17:50 +0100)]
with-authinfo-kludge.1: Let's start with a manpage.