<anarcat>
it looks like this system call never returns openat(AT_FDCWD, "/dev/tty", O_WRONLY|O_CLOEXEC)
boomboxnation has joined #foot
<dnkl>
I have a hard time seeing how a stalling syscall can be the terminals fault...
<anarcat>
that's such a weird bug
<anarcat>
but the reality is that it happens only in some terminals
<dnkl>
could be related somehow to the application doing different things depending on terminal? i.e using a different set of terminal features?
<anarcat>
i guess it's possible, but g10k doesn't do much with the terminal
<anarcat>
it might do colors and probe for that
chilledfrogs has quit [Quit: connection reset by purr]
<dnkl>
to my knowledge, openat can't not return. it smells now like a mutex deadlock, but I don't see how that could cause strace to think openat is stuck
<dnkl>
s/now/more
<dnkl>
interesting bug nevertheless
chilledfrogs has joined #foot
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
birdisword has quit [Ping timeout: 246 seconds]
birdisword has joined #foot
birdisword has quit [Ping timeout: 268 seconds]
moonshine has quit [Quit: .]
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
<Arnavion>
Maybe it's being opened multiple times and some circumstances cause the second one to block
<Arnavion>
If it really is stuck in openat then /proc/$pid/stack might give a hint
xgpt has joined #foot
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
birdisword has joined #foot
strom has quit [Remote host closed the connection]
strom has joined #foot
strom has quit [Read error: Connection reset by peer]
strom has joined #foot
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
ciara has quit [Remote host closed the connection]
ciara has joined #foot
boomboxnation has quit [Read error: Connection reset by peer]
boomboxnation has joined #foot
<ovf>
anarcat: in your strace, the openat completes two lines down. strace -f output is ordered by time, which means that traces of syscalls from different processes may interleave. this is what '<unfinished ...>' means -- the current syscall is still running but strace needs to print something else. it then continues with '<... $call resumed>'.
<ovf>
first token on a line of strace -f output is pid. you might also be interested in the -ff option if you find -f output confusing
<ovf>
dnkl: you were spot on here, but in general open can of course block. e.g. mkfifo foo;strace cat foo