the command tmux -CC fails to open any window but the process keeps running in the background...my iTerm2 version is 2.9.20150722 so I'm assuming its the latest
Designs
Child items
...
Show closed items
Linked items
0
Link issues together to show that they're related.
Learn more.
Strange, I've seen one other case like this were we write commands to tmux but never hear anything back. I'll add some more debug logging. If this remains a problem, please make a debug log with tomorrow's nightly build. Thanks.
It continues to be a problem on any Ubuntu 15.04 machine which has tmux 1.9. I can confirm this for this combination every time (5 machines tested) and Ubuntu 14.04 and tmux 1.8 works every time (3 machines tested).
I will check with the nightly you are referring to and report.
It is a total mystery. I have turned off the Mac built in fire wall and Little Snitch that I was running with no effect.
I am attaching the log and have also thought of one thing. My remote username does not match my local username. I use ssh to map it over (would that matter though?)
one more difference that I can think of...all my 15.04 machines have been upgraded from previous versions of Ubuntu...ok i reinstalled tmux (while using apt-get remove --purge to remove old configs...sadly no effect)
Are you able to build tmux? I made a fork of it and added some more debugging to the client log to see what the inputs and outputs are. Here are the steps:
Here is the most interesting thing. I have tried this repeatedly and it may be a race condition involved. your git repo compiled with debug enabled (--enable-debug option passed to configure) it is working repeatedly without any problems. But without the debug enabled, in -CC mode ...it works randomly ..maybe after first launch after compilation and then even after killing all sessions it does not (very bizarre and unpredictable). it does not reconnect to existing sessions at all without debug enabled.
I have thoroughly tested this multiple times.I also get same random behavior with tmux 2.0 stable (upstream) both with or without debug enabled.
So it appears (to me) that your additional debugging introduces some code change which makes it work (or so it seems)..hence my guess at a race condition.
Have attached logs with bot the flag enable-debug and otherwise from your git repo.
That's really good to know, and those logs confirm that something crazy is happening in tmux. We don't need it to be a debug build anyway, I can just keep adding more logging. Give me a few days because life is a touch chaotic right now. I'm quite keen to figure this one out.
Well, the debug build from your repo is the only one working (and I tried this twice since I couldn't believe the results) ...I'm happy to test with more logging :-) Although I hear great things about the tmux code base just haven't gotten around to digging in it
Nope. Also I have tried running tmux without my tmuxrc file to eliminate any issues coming out of that.
I think the clue lies in the fact that the debug build works perfectly (and only your github debug build) ...I have tried the official 2.0 sources...with and without debug and it doesn't work (it works once sometimes and then the next time onwards it does not)..but your github build is successfully working on every machine I have tried.
hey, no need to thank me. This is the open source contract of users helping developers (and those people who need developers to thank them for helping debug are idiots IMHO :) Thank you for sticking with this!
did you omit the --enable-debug flag? I am assuming it was for a reason and ran commands as you mentioned but here is the interesting part now ...for some reason it works once or twice and even reconnects (using attach) when it feels like and then it stops attaching and hangs.
I think this is the pattern. sometimes it does not attach to the existing process in which case it launches a new tmux process and it hangs
Attaching the 2 client and 1 server logs (for the time it connected first and then after detaching I tried to attach and it just hung)
Also to recover and create a new working session it is not sufficient to just do killall -9 tmux. I have to rm -rf /tmp/tmux-1000 folder also after which a new -CC tmux session gets created but whether it reattaches is upto chance (i.e. whether it creates a new one which hangs or attaches to the old one which works)
Note that your last build with debug enabled somehow always reattaches and works just fine
I updated my more_debugging branch. Please repeat the previous steps.
If you're curious (and so I don't forget):
I see that in the one that failed (the '947 log) newlines are ASCII 13 (CR), while in the other client log they're ASCII 10 (LF). This suggests that icrnl is off in the tty, since we send CR's but tmux expects LF's. However, tmux explicitly turns on icrnl when you run it with -CC. It's possible that tcsetattr is failing or that icrnl is somehow getting un-set.
I can confirm that the problem persists with tmux 2.0 which is the installed version in Ubuntu 15.10. it appears that versions after tmux 1.8 have an ongoing incompatibility. I will run your debugging branch and report again
tried the tmux 2.0 debug branch. the tmux integration works fine randomly with this branch! Again I can't seem to understand the pattern. What I can report for sure is the it only works with the -vvvvvvv option, never without it and that too only in the debug branch.
I would say, I am comfortable following directions :-) ..I have used pdb and other debuggers so I am familiar with debugging but not those tools but can't be that different to play along.