Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: There are two problems here. One (the big problem), I made the mistake of changing from flock() to lockf() which is badly designed and next to useless. Two, there is a small window where two clients could race and both create a server. So please try this which changes back to flock, fixes the race and adds some logging. Index: client.c Did you direct this post to the tmux community in general or are you asking me to recompile tmux and try the same scenario? If you are, I'm most willing to do that, of course, but you'll have to provide me with some guidelines. I suppose the changes you made isn't yet in my distro's repos (for any release), so I'd have to get the source, apply the patch (if it isn't already), and then recompile. If someone else feels like doing this the zsh init file that brought along the problem is: http://user.it.uu.se/~embe8573/conf/.zprofile -- underground experts united: http://user.it.uu.se/~embe8573 -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Yes, you will need to get the source from git then # patch ~/saved-email # sh autogen.sh # ./configure # make Then copy the tmux binary to replace your existing one. On Wed, Jun 25, 2014 at 05:05:33AM +0200, Emanuel Berg wrote: Nicholas Marriott nicholas.marri...@gmail.com writes: There are two problems here. One (the big problem), I made the mistake of changing from flock() to lockf() which is badly designed and next to useless. Two, there is a small window where two clients could race and both create a server. So please try this which changes back to flock, fixes the race and adds some logging. Index: client.c Did you direct this post to the tmux community in general or are you asking me to recompile tmux and try the same scenario? If you are, I'm most willing to do that, of course, but you'll have to provide me with some guidelines. I suppose the changes you made isn't yet in my distro's repos (for any release), so I'd have to get the source, apply the patch (if it isn't already), and then recompile. If someone else feels like doing this the zsh init file that brought along the problem is: http://user.it.uu.se/~embe8573/conf/.zprofile -- underground experts united: http://user.it.uu.se/~embe8573 -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: Also please run all the tmux with - without the sleeps and see what is in the tmux-client-* logs that should end up in your home directories. I run it twice as first time it worked without the sleeps. Then it didn't. Here are the logs for both cases: http://user.it.uu.se/~embe8573/tmux-logs/ -- underground experts united: http://user.it.uu.se/~embe8573 -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: Hmm. One of the clients should win and start the server, you shouldn't end up with three tmux servers. Are you moving tmux sockets away from /tmp? No, not that I know of. How would I do that? (Then I could tell you instantly if I did it.) If not, what filesystem is /tmp? How do I check that, with fsck? $ fsck -N /tmp fsck from util-linux 2.20.1 [/sbin/fsck.ext2 (1) -- /tmp] fsck.ext2 /tmp -- underground experts united: http://user.it.uu.se/~embe8573 -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: Also please run all the tmux with - without the sleeps and see what is in the tmux-client-* logs that should end up in your home directories. Oops, I forgot this line in ~/.xinitrc - xterm -fullscreen -e tmux new-session\; split-window -v\; select-pane -U - but that tmux (in xterm) wasn't a failing one so perhaps I don't need to do it over again? All the other tmux invocations (in ~/.zprofile) had -'s. By the way, is it possible to troubleshot this not having to restart the computer, but still get the same behaviour and order of execution of dotfiles etc.? That'd be less disruptive... -- underground experts united: http://user.it.uu.se/~embe8573 -- Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). Without (any) sleep, instant failure for /dev/tty2 and /dev/tty3: pid ppid comm -- 29821 tmux 29831 tmux 29841 tmux -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). Using this insight, this will do unless you guys have better suggestions: tmux-two-panes-50-50 () { tmux new-session\; split-window -v\; select-pane -U } wait-for-tmux () { while [[ `ps -e -o ppid,comm | grep '1 tmux' | wc -l` != 1 ]] ; do sleep 1 done } case $TTY in (/dev/tty1) emacs ;; (/dev/tty2) tmux-two-panes-50-50 ;; (/dev/tty3) wait-for-tmux; tmux-two-panes-50-50 ;; (/dev/tty4) wait-for-tmux; tmux new-session\; split-window -v\; resize-pane -D 10 ;; (/dev/tty5) wait-for-tmux; start_X_if_not_running; irssi ;; esac -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Romain Francoise rfranco...@debian.org writes: It's not possible to have per-pane or per-window key bindings, but you can just start the application in a dedicated tmux server (using e.g. 'tmux -L rebind -c rebind.conf') and then attach to that from the main server. In the inner server you can disable any status lines and use another prefix key. I didn't get this to work. Should I put the tmux bind stuff in the rebind.conf file, as well as the command to launch the application? That works, but it is tmux global. If I attach with 'tmux attach' I only see the same thing - what do I do with it? -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Romain Francoise rfranco...@debian.org writes: It's not possible to have per-pane or per-window key bindings, but you can just start the application in a dedicated tmux server (using e.g. 'tmux -L rebind -c rebind.conf') and then attach to that from the main server. In the inner server you can disable any status lines and use another prefix key. OK, got this to work as well - that rebind.conf and attach made me confused, but it is actually very simple: In a tmux pane, this tmux -L rebind bind x -n send-keys q won't affect the other panes. Thank you! -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Romain Francoise rfranco...@debian.org writes: Anyway, I have no idea about this but my advice would be to take the reverse approach and use tmux as your shell, a new session will automatically be started for each tty. I just tried that, but I rather not do that. First, I want Emacs in tty1 without tmux because I have shortcuts in the console, in Emacs, and in tmux, and they are setup during a huge (but pleasant) effort, and they immediately broke (of course). Also, I don't won't Emacs on top of tmux because I don't need it. Though I suspect the overhead is slim, it also (just an example) changed my block cursor which I like in Emacs, to the underscore which I like in tmux/zsh. Also, even though I did try to run zsh from tmux, many commands wouldn't work (such basic things as shutdown, for example). I realize this has to do with my elaborate setup and is solvable, but I rather not do it because I'm happy the way it is, safe for the strange bug (?) I just mentioned. -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: What tmux version? $ tmux -V tmux 1.9 If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). OK, I'll remove the sleeps and do that the next time it happens. As for now the only test I did was tmux list-clients and then I saw that in the failing tty, that tty/client wasn't listed. -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Nicholas Marriott nicholas.marri...@gmail.com writes: If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). Now I get it! I reinserted the sleepyheads and it works, this time I get $ ps -e -o pid,ppid,comm | grep tmux 2979 2969 tmux 2983 1 tmux 2990 2968 tmux 3017 2940 tmux 3050 3042 tmux I.e., only one server! That is, when I start all tmuxes at almost the same time (without sleep), all of them feel they are the first, as the others haven't initialized at that point. Man, I should be a private eye... Thanks Mr. Marriott for the clue! Nonetheless, there should be a better way than sleep to do it. I'm on it. PS. I hope it is not a major annoyance having to confirm my posts. If you like, feel free to add me to the list so you won't have to do that. DS. -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Hmm. One of the clients should win and start the server, you shouldn't end up with three tmux servers. Are you moving tmux sockets away from /tmp? If not, what filesystem is /tmp? On Sun, Jun 22, 2014 at 04:50:00AM +0200, Emanuel Berg wrote: Nicholas Marriott nicholas.marri...@gmail.com writes: If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). Without (any) sleep, instant failure for /dev/tty2 and /dev/tty3: pid ppid comm -- 29821 tmux 29831 tmux 29841 tmux -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Also please run all the tmux with - without the sleeps and see what is in the tmux-client-* logs that should end up in your home directories. On Mon, Jun 23, 2014 at 10:43:53AM +0100, Nicholas Marriott wrote: Hmm. One of the clients should win and start the server, you shouldn't end up with three tmux servers. Are you moving tmux sockets away from /tmp? If not, what filesystem is /tmp? On Sun, Jun 22, 2014 at 04:50:00AM +0200, Emanuel Berg wrote: Nicholas Marriott nicholas.marri...@gmail.com writes: If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). Without (any) sleep, instant failure for /dev/tty2 and /dev/tty3: pid ppid comm -- 29821 tmux 29831 tmux 29841 tmux -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
On Sun, Jun 22, 2014 at 10:50:51PM +0200, Emanuel Berg wrote: Nonetheless, there should be a better way than sleep to do it. I'm on it. tmux is supposed to do this the right way since 1.7, using a lockfile. Your experience suggests that it's not working. -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
for a given value of right - it should work if you aren't do anything weird code looks ok to me but clearly it's broken somewhere On Mon, Jun 23, 2014 at 11:33:29AM +0200, Romain Francoise wrote: On Sun, Jun 22, 2014 at 10:50:51PM +0200, Emanuel Berg wrote: Nonetheless, there should be a better way than sleep to do it. I'm on it. tmux is supposed to do this the right way since 1.7, using a lockfile. Your experience suggests that it's not working. -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
Romain Francoise rfranco...@debian.org writes: The mailing list is tmux-us...@lists.sf.net and is usually pretty active. OK! I send this mail to them as well, as well as the first mail I sent, in quotes. I don't know if I'm added automatically but otherwise you'll have to CC me, I guess. In .zprofile, I have a setup like this to start tmux in a few of the ttys: tmux-two-panes-50-50 () { tmux new-session\; split-window -v\; select-pane -U } # ... case $TTY in (/dev/tty1) emacs ;; (/dev/tty2) sleep 4; tmux-two-panes-50-50 ;; (/dev/tty3) sleep 6; tmux-two-panes-50-50 ;; (/dev/tty4) sleep 8; tmux new-session\; split-window -v\; resize-pane -D 10 ;; (/dev/tty5) start_X_if_not_running; irssi ;; esac Also, in, .xinitrc, the last line is xterm -fullscreen -e tmux new-session\; split-window -v\; select-pane -U so I get tmux there as well. The problem with this is, sometimes (once in ten boots perhaps) 1-2 of the clients disappear. tmux actually appears in the failing tty(s), but it is somehow crippled (e.g., 'tmux select-pane -U' doesn't work etc.). And, with 'tmux list-clients' there isn't an entry for that tty! If you notice the 'sleep' commands in .zprofile, I have no idea why but those have a positive effect as now this problem is much more rare, and it always affects a single tty, never two (or more). But you're not logging in on all the ttys at once, are you? Well, when I start Debian, in /etc/inittab, I have the following: 1:2345:respawn:/sbin/getty -a ... 38400 tty1 2:23:respawn:/sbin/getty -a ... 38400 tty2 3:23:respawn:/sbin/getty -a ... 38400 tty3 4:23:respawn:/sbin/getty -a ... 38400 tty4 5:23:respawn:/sbin/getty -a ... 38400 tty5 So in that sense, logging should be pretty close. But: the .zprofile file must be read for every tty otherwise how can the case statement give different behaviour what to do, depending on what tty it is? So it should be individual invocations of tmux as well, right? But (especially without sleep) sometimes one or two clients don't appear. Why would the sleep calls change anything? I don't know except I know that it helps, a lot. Anyway, I have no idea about this but my advice would be to take the reverse approach and use tmux as your shell, a new session will automatically be started for each tty. Yeah, you mean change shell as in chsh? Yes, /usr/bin/tmux is in /etc/shells so I'll try that at once! You lose the ability to configure panes differently for each tty, but your setup looks simple enough that it could be done some other way. Yes. How can I use for example tmux bind -n i send-keys Up but have it only affect the pane/window/client where it is invoked? I'd like to set one tty apart for an ncurses application, which cannot be key-configured. So I thought I'd use tmux to translate my keystrokes to those of the application. That works, however, that also breaks the keyboards for all other panes, in all other ttys (and in X) where tmux runs. Ideas? It's not possible to have per-pane or per-window key bindings, but you can just start the application in a dedicated tmux server (using e.g. 'tmux -L rebind -c rebind.conf') and then attach to that from the main server. In the inner server you can disable any status lines and use another prefix key. OK, I didn't understand that but I'll try it. -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users
Re: tmux bug (?) and problem (and more)
What tmux version? If you do without the sleeps and it goes wrong, how many tmux servers do you end up with? (run ps -eopid,ppid,comm and look for tmux with a ppid of 1). On Sat, Jun 21, 2014 at 11:16:18PM +0200, Emanuel Berg wrote: Romain Francoise rfranco...@debian.org writes: The mailing list is tmux-us...@lists.sf.net and is usually pretty active. OK! I send this mail to them as well, as well as the first mail I sent, in quotes. I don't know if I'm added automatically but otherwise you'll have to CC me, I guess. In .zprofile, I have a setup like this to start tmux in a few of the ttys: tmux-two-panes-50-50 () { tmux new-session\; split-window -v\; select-pane -U } # ... case $TTY in (/dev/tty1) emacs ;; (/dev/tty2) sleep 4; tmux-two-panes-50-50 ;; (/dev/tty3) sleep 6; tmux-two-panes-50-50 ;; (/dev/tty4) sleep 8; tmux new-session\; split-window -v\; resize-pane -D 10 ;; (/dev/tty5) start_X_if_not_running; irssi ;; esac Also, in, .xinitrc, the last line is xterm -fullscreen -e tmux new-session\; split-window -v\; select-pane -U so I get tmux there as well. The problem with this is, sometimes (once in ten boots perhaps) 1-2 of the clients disappear. tmux actually appears in the failing tty(s), but it is somehow crippled (e.g., 'tmux select-pane -U' doesn't work etc.). And, with 'tmux list-clients' there isn't an entry for that tty! If you notice the 'sleep' commands in .zprofile, I have no idea why but those have a positive effect as now this problem is much more rare, and it always affects a single tty, never two (or more). But you're not logging in on all the ttys at once, are you? Well, when I start Debian, in /etc/inittab, I have the following: 1:2345:respawn:/sbin/getty -a ... 38400 tty1 2:23:respawn:/sbin/getty -a ... 38400 tty2 3:23:respawn:/sbin/getty -a ... 38400 tty3 4:23:respawn:/sbin/getty -a ... 38400 tty4 5:23:respawn:/sbin/getty -a ... 38400 tty5 So in that sense, logging should be pretty close. But: the .zprofile file must be read for every tty otherwise how can the case statement give different behaviour what to do, depending on what tty it is? So it should be individual invocations of tmux as well, right? But (especially without sleep) sometimes one or two clients don't appear. Why would the sleep calls change anything? I don't know except I know that it helps, a lot. Anyway, I have no idea about this but my advice would be to take the reverse approach and use tmux as your shell, a new session will automatically be started for each tty. Yeah, you mean change shell as in chsh? Yes, /usr/bin/tmux is in /etc/shells so I'll try that at once! You lose the ability to configure panes differently for each tty, but your setup looks simple enough that it could be done some other way. Yes. How can I use for example tmux bind -n i send-keys Up but have it only affect the pane/window/client where it is invoked? I'd like to set one tty apart for an ncurses application, which cannot be key-configured. So I thought I'd use tmux to translate my keystrokes to those of the application. That works, however, that also breaks the keyboards for all other panes, in all other ttys (and in X) where tmux runs. Ideas? It's not possible to have per-pane or per-window key bindings, but you can just start the application in a dedicated tmux server (using e.g. 'tmux -L rebind -c rebind.conf') and then attach to that from the main server. In the inner server you can disable any status lines and use another prefix key. OK, I didn't understand that but I'll try it. -- underground experts united: http://user.it.uu.se/~embe8573 -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users -- HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions Find What Matters Most in Your Big Data with HPCC Systems Open Source. Fast. Scalable. Simple. Ideal for Dirty Data. Leverages Graph Analysis for Fast Processing Easy Data Exploration http://p.sf.net/sfu/hpccsystems ___ tmux-users mailing list tmux-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tmux-users