This turns out to be something pretty simple - we should not be
executing commands in a client context before that client has finished
identifying (that is, before we have the received the environment):
diff --git a/server.c b/server.c
index e9c2298..1fbaf37 100644
--- a/server.c
+++ b/server.c
@@ -196,8 +196,10 @@ server_loop(void)
do {
items = cmdq_next(NULL);
- TAILQ_FOREACH(c, &clients, entry)
- items += cmdq_next(c);
+ TAILQ_FOREACH(c, &clients, entry) {
+ if (c->flags & CLIENT_IDENTIFIED)
+ items += cmdq_next(c);
+ }
} while (items != 0);
server_client_loop();
On Wed, Nov 09, 2016 at 12:36:47AM +0100, Frank Terbeck wrote:
> Hey list,
>
> I recently upgraded tmux to its latest git master tip. I am starting my
> tmux server by using "tmux start-server" with my configuration file
> starting the server's initial session (a couple of shells, mostly).
>
> After attaching, none of the shells have a DISPLAY environment variable
> set. "tmux show-environment -g" and "tmux show-environment" both have
> this variable set, and indeed, it *is* set in the process that spawns
> the start-server call.
>
> So, I've taken a look at tmux's revision history. Bisecting isn't quite
> as easy as one might think, since you can only build commits on the
> master branch of the portable version of the program. But I still think
> I could narrow it down to a single change set:
>
> * commit c67b7025882ee6bc2b3d5112dbc630c191f1fa2a
> |\ Merge: 3d8efdf ddc4512
> | | Author: Thomas Adam <[email protected]>
> | | Date: Sun Oct 16 20:01:10 2016 +0100
> | |
> | | Merge branch 'obsd-master'
> | |
> | * commit ddc4512d2e0eda6c705e002cb5dbf80719d709e1
> | | Author: nicm <nicm>
> | | Date: Sun Oct 16 17:55:14 2016 +0000
> | |
> | | Rewrite command queue handling. [...]
> | |
> * | commit 3d8efdf310f475a6b9ed83814182c966de2bdab3
> |\ Merge: a15939b bfe14b5
> | | Author: Thomas Adam <[email protected]>
> | | Date: Sun Oct 16 02:01:11 2016 +0100
> | |
> | | Merge branch 'obsd-master'
>
> Building c67b7025 shows the faulty behavior, while 3d8efdf31 is fine,
> which suggests that ddc4512d2e0 is the change set that introduces the
> breakage.
>
> Here is my recipe for reproducing the issue: With a file "foobar"
> containing this: "new-session -d -s misc", I am doing this with each
> build:
>
> % echo $DISPLAY
> :0
> % git checkout c67b7025
> % ./autogen.sh && ./configure && make
> % ./tmux -f ~/foobar -L test start-server
> % ./tmux -L test attach-session
> [..in that session..]
> % echo $DISPLAY
> % exit
> [..this closes the server, back to the parent shell..]
> % git checkout 3d8efdf31
> [..repeat recipe with this state..]
>
> In the c67b7025 build's session, $DISPLAY is unset, while in the
> 3d8efdf31 build, it is set to ":0".
>
> Unfortunately, ddc4512 is a rather substantial change set, which I've
> skimmed through, but I can't quite tell what in it might cause this. I
> am sure Nicholas or Thomas will have an easier time. :)
>
>
> Regards, Frank
> --
> In protocol design, perfection has been reached not when there is
> nothing left to add, but when there is nothing left to take away.
> -- RFC 1925
>
> --
> You received this message because you are subscribed to the Google Groups
> "tmux-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To post to this group, send an email to [email protected].
> For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups
"tmux-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send an email to [email protected].
For more options, visit https://groups.google.com/d/optout.
