On 5/14/20 8:34 PM, Jan Gorecki wrote:> Hi R developers,
>
> I observed that system(timeout=) may still return exit code 0, when
> killing the process due to timeout.
>
> In src/unix/sys-unix.c there is
>
> #define KILL_SIGNAL1 SIGINT
> #define KILL_SIGNAL2 SIGTERM
> #define KILL_SIGNAL3 SIGKILL
> #define EMERGENCY_TIMEOUT 20
>
> After little bit of debugging I observed that total time of system
> call is provided timeout value + 20s. That means EMERGENCY_TIMEOUT 20
> kicked in, adding 20 seconds.
>
> I don't have a reproducible example, but following code, and output
> file below should be enough to ensure that there is a problem there
> (exit code 0 despite timeout).
Thanks for the report, but I can't tell from the report what the problem
is, I would really need a reproducible example, and ideally one that
uses only minimal/trivial child processes. I tried with such examples on
my system, but they did not trigger the problem.
In this example, SIGINT succeeds in terminating the child process
> system.time(print(system("while /bin/true ; do /bin/true ;
done",
timeout=1)))
[1] 124
?? user? system elapsed
? 0.712?? 0.316?? 1.002
In this example, SIGTERM succeeds (after additional 20s)
> system.time(print(system("trap '' INT ; while /bin/true ; do
/bin/true ; done", timeout=1)))
[1] 124
?? user? system elapsed
?14.813?? 6.641? 21.002
In this example, SIGKILL succeeds (after additional 40s)
> system.time(print(system("trap '' INT TERM ; while /bin/true
; do
/bin/true ; done", timeout=1)))
[1] 124
?? user? system elapsed
?29.075? 12.918? 41.002
In these examples, the execution time and the exit status is as
documented in ?system.
Best
Tomas
>
> warn = NULL
> p = proc.time()[[3L]]
> tryCatch(
> ret <- system(cmd, timeout=timeout_s),
> warning = function(w) {
> warn <<- w[["message"]]
> }
> )
> if (length(warn) && ret==0L)
> cat(sprintf("command '%s' timed out(?) but still exited
with 0 code,
> timeout %ss, took %ss, warning '%s'\n",
> cmd, timeout_s, proc.time()[[3L]]-p, warn),
> file="timeout-exit-codes.out", append=TRUE)
>
> And the content of timeout-exit-codes.out:
>
> command '/bin/bash -c "./_launcher/solution.R > log.out 2>
log.err"'
> timed out(?) but still exited with 0 code, timeout 7200s, took
> 7220.005s, warning '/bin/bash -c "./_launcher/solution.R >
log.out 2>
> log.err"' timed out after 7200s'
>
> Thank you,
> Jan Gorecki
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel