os/exec: TestNohup fails inside tmux

[petar]

darwin
amd64
08d20469cc20 tip


When running ./all.bash inside a tmux session, one of the tests fails.

When running it from the "normal" console (in Terminal), it passes fine.

[davecheney]

Comment 1:

Please provide some additional details, eg, which test fails, and its output when
failing.

Status changed to WaitingForReply.

[petar]

Comment 2:

Oh, apologies, forgot. Here is the tail of the ./all.bash print out:
ok      net/http/httptest       0.035s
ok      net/http/httputil       0.036s
?       net/http/pprof  [no test files]
ok      net/mail        0.020s
ok      net/rpc 0.062s
ok      net/rpc/jsonrpc 0.039s
ok      net/smtp        0.030s
ok      net/textproto   0.029s
ok      net/url 0.033s
ok      os      0.058s
ok      os/exec 0.609s
--- FAIL: TestNohup-10 (0.13 seconds)
        signal_test.go:205: ran test with -send_uncaught_sighup=1
under nohup and it failed: expected success.
                Error: exit status 127
                Output:
                nohup: can't detach from console: Inappropriate ioctl for device
FAIL
FAIL    os/signal       0.323s
ok      os/user 0.024s
ok      path    0.018s
ok      path/filepath   0.508s
ok      reflect 0.045s
ok      regexp  0.106s
ok      regexp/syntax   0.648s
ok      runtime 3.126s
?       runtime/cgo     [no test files]
ok      runtime/debug   0.026s
ok      runtime/pprof   0.340s
ok      sort    0.066s
ok      strconv 0.345s
ok      strings 0.126s
ok      sync    0.070s
ok      sync/atomic     0.029s
ok      syscall 0.044s
?       testing [no test files]
?       testing/iotest  [no test files]
ok      testing/quick   0.035s
ok      text/scanner    0.021s
ok      text/tabwriter  0.021s
ok      text/template   0.045s
ok      text/template/parse     0.024s
ok      time    2.354s
ok      unicode 0.021s
ok      unicode/utf16   0.021s
ok      unicode/utf8    0.019s
?       unsafe  [no test files]

[davecheney]

Comment 3:

Thanks for the additional detail. I think there is a known issue with SIGHUP, or at
least who owns it. This issue may be a duplicate of an existing one, i'll let someone
else decide.

Status changed to Accepted.

[balasanjay]

Comment 4:

I also saw this failure, but I accidentally found a solution when fixing another tmux
issue.
Try using https://github.com/ChrisJohnsen/tmux-MacOSX-pasteboard
That fixed it for me (I think its some OSX weirdness, but didn't investigate in detail).

[davecheney]

Comment 5:

I've had confirmations of this bug, again tmux and osx. The workaround is clear, don't
build go inside tmux.

Labels changed: added priority-later, removed priority-triage.

[rsc]

Comment 7:

Labels changed: added go1.2maybe.

[robpike]

Comment 8:

Deferring to 1.3. A peculiarity of the execution environment.

Labels changed: added go1.3maybe, removed go1.2maybe.

[robpike]

Comment 9:

Labels changed: removed go1.3maybe.

[minux]

Comment 10:

when trying to restoring the netbsd-386 builder, i also encountered this issue,
so it's not a OS X specific issue.
http://build.golang.org/log/d64e9b65e5e39f47982df8dfc28f2f25fa270813
i think we should fix it before we release Go 1.2 so that it could be tested on
all supported platforms. Opinions?

[minux]

Comment 11:

update: running all.bash inside screen(1) is ok.

[rsc]

Comment 12:

Labels changed: added go1.3maybe.

[rsc]

Comment 13:

Labels changed: added release-none, removed go1.3maybe.

[rsc]

Comment 14:

Labels changed: added repo-main.

[jacobsa]

FWIW, this still happens to me when attempting to build Go 1.5 in tmux 2.0 on OS X 10.10.5. Does anybody understand exactly what's going on?

[ianlancetaylor]

I wouldn't say I know exactly what's going on, but the broad details seem clear enough: you can't run /usr/bin/nohup on a tmux. If you do, it will fail with the inappropriate ioctl error message. You should be able to create this yourself in a tmux: "/usr/bin/nohup ls".

What we should do in Go is recognize that error message and, if we see it, skip the test rather than failing it. I don't use Darwin so I can't recreate the problem myself (tmux on GNU/Linux works fine), but it ought to be an easy fix for anyone who can.

[davecheney]

Googling results are inconclusive

http://thread.gmane.org/gmane.comp.terminal-emulators.tmux.user/4450

It seems that nobody really cares about this use case on darwin, well, from
the tmux side at least.

Tmux makes itself known with env vars, perhaps I can skip the test if we
see those env vars.

On Tue, Aug 25, 2015 at 3:24 AM, Ian Lance Taylor [email protected]
wrote:

I wouldn't say I know exactly what's going on, but the broad details seem
clear enough: you can't run /usr/bin/nohup on a tmux. If you do, it will
fail with the inappropriate ioctl error message. You should be able to
create this yourself in a tmux: "/usr/bin/nohup ls".

What we should do in Go is recognize that error message and, if we see it,
skip the test rather than failing it. I don't use Darwin so I can't
recreate the problem myself (tmux on GNU/Linux works fine), but it ought to
be an easy fix for anyone who can.

—
Reply to this email directly or view it on GitHub
#5135 (comment).

[jacobsa]

@davecheney: Are you looking at this further? If not, I'll send a CL.

[davecheney]

I am not actively looking at this at the moment.

On Tue, 25 Aug 2015 08:25 Aaron Jacobs [email protected] wrote:

@davecheney https://github.com/davecheney: Are you looking at this
further? If not, I'll send a CL.

—
Reply to this email directly or view it on GitHub
#5135 (comment).

[jacobsa]

Okay, I'll prepare a CL.

For the record, the issue is misnamed: the failing test is in os/signal.

[gopherbot]

CL https://golang.org/cl/13883 mentions this issue.