I think the problem is that you're expecting "$LINENO"
to give you the line of execution for the last command, which might almost work, but clean_a()
also gets its own $LINENO
and that you should do instead:
error "something!
line: $1
...
But even that probably wouldn't work because I expect it will just print the line on which you set the trap
.
Here's a little demo:
PS4='DEBUG: $LINENO : ' \
bash -x <<\CMD
trap 'fn "$LINENO"' EXIT
fn() { printf %s\\n "$LINENO" "$1"; }
echo "$LINENO"
CMD
OUTPUT
DEBUG: 1 : trap 'fn "$LINENO"' EXIT
DEBUG: 3 : echo 3
3
DEBUG: 1 : fn 1
DEBUG: 2 : printf '%s\n' 2 1
2
1
So the trap
gets set, then, fn()
is defined, then echo
is executed. When the shell completes executing its input, the EXIT
trap is run and fn
is called. It is passed one argument - which is the trap
line's $LINENO
. fn
prints first its own $LINENO
then its first argument.
I can think of one way you might get the behavior you expect, but it kinda screws up the shell's stderr
:
PS4='DEBUG: $((LASTNO=$LINENO)) : ' \
bash -x <<\CMD
trap 'fn "$LINENO" "$LASTNO"' EXIT
fn() { printf %s\\n "$LINENO" "$LASTNO" "$@"; }
echo "$LINENO"
CMD
OUTPUT
DEBUG: 1 : trap 'fn "$LINENO" "$LASTNO"' EXIT
DEBUG: 3 : echo 3
3
DEBUG: 1 : fn 1 3
DEBUG: 2 : printf '%s\n' 2 1 1 3
2
1
1
3
It uses the shell's $PS4
debug prompt to define $LASTNO
on every line executed. It's a current shell variable which you can access anywhere within the script. That means that no matter what line is currently being accessed, you can reference the most recent line of the script run in $LASTNO
. Of course, as you can see, it comes with debug output. You can push that to 2>/dev/null
for the majority of the script's execution maybe, and then just 2>&1
in clean_a()
or something.
The reason you get 1
in $LASTNO
is because that is the last value to which $LASTNO
was set because that was the last $LINENO
value. You've got your trap
in the archieve_it()
function and so it gets its own $LINENO
as is noted in the spec below. Though it doesn't appear that bash
does the right thing there anyway, so it may also be because the trap
has to re-exec the shell on INT
signal and $LINENO
is therefore reset. I'm a little fuzzy on that in this case - as is bash
, apparently.
You don't want to evaluate $LASTNO
in clean_a()
, I think. Better would be to evaluate it in the trap
and pass the value trap
receives in $LASTNO
through to clean_a()
as an argument. Maybe like this:
#!/bin/bash
PS4='^MDEBUG: $((LASTNO=$LINENO)) : '; set -x
archieve_it () {
trap 'clean_a $LASTNO $LINENO "$BASH_COMMAND"' \
SIGHUP SIGINT SIGTERM SIGQUIT
while :; do sleep 1; done
} 2>/dev/null
clean_a () { : "$@" ; } 2>&1
Try that - it should do what you want, I think. Oh - and note that in PS4=^M
the ^M
is a literal return - like CTRL+V ENTER.
From the POSIX shell spec:
Set by the shell to a decimal number representing the current sequential line number (numbered starting with 1) within a script or function before it executes each command. If the user unsets or resets LINENO
, the variable may lose its special meaning for the life of the shell. If the shell is not currently executing a script or function, the value of LINENO
is unspecified. This volume of IEEE Std 1003.1-2001 specifies the effects of the variable only for systems supporting the User Portability Utilities option.
I've observed that using CtrlZ to shift the program to a background process does the trick.
foo && bar
Thanks to @Arkadiusz Drabczyk for pointing it in comments that foo; bar
doesn't give control in the required way.
Then:
^Z
[1]+ Stopped foo
The command stops only the first task and
fg %1
This brings only task foo
to the foreground and completes the task and exits.
PS: This can be checked with two scripts writing to a file. The first one sleeping for a few seconds to give time to be put back.
I'm lost on why the CtrlZ handles only the command running and leaves the rest. Would love to get to know.
Best Answer
It should work if you just
trap
SIGINT
to something. Like:
(true
).Interrupting the
something...
doesn't make the shell exit now, since it ignores the signal. However, if you ^C thesleep
process, it will exit with a failure, and the loop stops due to that. Move thesleep
to the inside of the loop or add something like|| true
to prevent that.Note that if you use
trap "" INT
to ignore the signal completely (instead of assigning a command to it), it's also ignored in the child process, so then you can't interruptsomething...
either. This is explicitly mentioned in at least Bash's manual: