Race condition hangs on multiple mintty/tcsh? Brad Wetmore

[Thomas Wolff]

Am 06.08.2020 um 20:07 schrieb Brad Wetmore via Cygwin:
> [I apologize if this doesn't thread correctly:  I've been getting the digest, and can't respond directly.  And I have no idea how my name ended up in the Subject line:  Rookie Move!]
>
> More debug information, and a workaround/solution:
>
> I am running:
>
>      Windows 10 Version 1909, fully patched.
>      Cygwin v3.1.6-1
>      Openssh v8.3p1-1
>      mintty v3.2.0-1
>      tcsh v6.22.02-1
>
> The issue seems to be a race condition setting up ssh-agent, and then immediately starting up the the mintty/tcsh terminals. My guess is that something in ssh-agent is not fully setup yet, and is causing read hiccups in mintty.
>
> My startup script looks like:
>
>       #!/bin/sh
>
>      if [ -z "$SSH_AGENT_PID" ] ; then
>           echo "Starting ssh-agent"
>           eval $(ssh-agent)
>           ssh-add
>           #sleep 3
>       fi
>
>      C:/cygwin64/bin/mintty.exe -i /usr/share/mintty/icon/wsl.ico \
>           --position 0,0 --size 80x24 --fg black --bg white /bin/tcsh &
>       #sleep 1
>       ...9 more invocations deleted...
>
> Without the sleep 3 after starting ssh-agent, I see one of three conditions:
>
> 1. The mintty window is created, but immediately crashes printing a dozen or so "Use "exit" to leave tcsh" before it does. At least that I think it says, it happens fast! ;)
>
> 2. The mintty window is created, and is ready for input. On the first keypress, it prints a dozen or so "Use "exit" to leave tcsh" and crashes.
>
> 3. The mintty window is created, and hangs. No prompt.
>
> I get conditions #1/2 when I don't pause between mintty invocations.
> I get condition #2 when I pause for 1 second between mintty invocations.
>
> If I pause for 3 seconds after starting ssh-agent, then everything works as expected.
>
> Note this issue occurs only during the first setup of 10 terminals. Running it again does not tickle the issue, probably because ssh-agent is fully up.
>
> Hope this helps.
You had previously reported the issue without ssh involved. Does this 
make it reproducible more often? Otherwise let's not make things 
unnecessarily complicated.