Define bash function which does not show up in xtrace (set -x)
You cannot change how the function is invoked from inside the function, but you could define the function to invoke a subshell which then immediately turns off tracing; notice the parenthesis surrounding the body instead of the typical braces:
log() (
set +x
# rest of log()
)
Invoking log()
then generates only the call itself (from the existing set -x
code) and the set +x
call, not the rest of the subsequent commands in the function. The existing set -x
setting is restored once the function exits.
A quick and dirty hack that should work in all shells is to (temporarily) make your log
an external script instead of a function.
In bash you could also use a trap '...' DEBUG
and shopt -s extdebug
combination, which is much more versatile than set -x
. Example:
debug() {
local f=${FUNCNAME[1]} d=${#FUNCNAME[@]} c=$BASH_COMMAND
if [ "$NOTRACE" ]; then
case $debug_skip in ''|$d) debug_skip=;; *) return;; esac
eval "case \$c in $NOTRACE) debug_skip=\$d; return; esac"
fi
# before the 1st command in a function XXX
case $c in $f|"$f "*) return;; esac
printf >&2 "%*s(%s) %s\n" $((d * 2 - 4)) "" "$f" "$c"
}
(Of course, you can dispense with the strange format + indenting and make it completely set-x
-like; you can also redirect it to another file instead of mixed with the stderr from commands.)
Then:
NOTRACE='"log "*'
shopt -s extdebug
trap debug DEBUG
log(){ log1 "$@"; }; log1(){ log2 "$@"; }
log2(){ log3 "$@"; }; log3(){ echo "$@"; }
foo(){ foo1 "$@"; }; foo1(){ foo2 "$@"; }
foo2(){ foo3 "$@"; }; foo3(){ echo "$@"; }
bar(){ case $# in 0) ;; *) echo "$1"; shift; bar "$@";; esac; }
foo 1 2 3
log 7 8 9
bar 1 2 3 4
will result in:
(main) foo 1 2 3
(foo) foo1 "$@"
(foo1) foo2 "$@"
(foo2) foo3 "$@"
(foo3) echo "$@"
1 2 3
7 8 9
(main) bar 1 2 3 4
(bar) case $# in
(bar) echo "$1"
1
(bar) shift
(bar) case $# in
(bar) echo "$1"
2
(bar) shift
(bar) case $# in
(bar) echo "$1"
3
(bar) shift
(bar) case $# in
(bar) echo "$1"
4
(bar) shift
(bar) case $# in