Lets assume the following situation:
function add_to_list
{
thelist="$thelist $$"
}
for i in $(seq 1 100); do
add_to_list&
done
wait
echo "The list: $thelist"
This obviously does not work as it should, because the threads are accessing the same variable - how should it be done properly
Bash does not support threads. Only subprocesses. And it is impossible to change a variable of parent process in a subprocess.
You might use a shared storage that handle concurrent updates, like a database.
Here is an example based on sqlite3:
function create_table
{
sqlite3 thelist <<-%
drop table t;
create table t(one varchar(255));
%
}
function add_to_list
{
sqlite3 thelist <<-%
.timeout 60000
insert into t values('$$');
%
}
function dump_list
{
sqlite3 thelist <<-%
select * from t;
%
}
create_table
for i in $(seq 1 100); do
add_to_list&
done
wait
thelist="$(dump_list)"
echo "The list: $thelist"
It seems to me that you're conflating threading and subprocessing semantics, especially wanting a variable to be updated.
You could use a named pipe like to communicate between the processes.
$ mkfifo mypipe
$ cat < mypipe
$ seq 1 10 > mypipe
This makes cat print the sequence.
bash-3.2$ cat always.sh
while : ; do
cat < mypipe
done
bash-3.2$ seq 1 100 > mypipe & seq 100 200 > mypipe & seq 200 300 > mypipe &
Creates out from always.sh 1 to 300 in order.
I'm unsure whether multiple processes writing to the same named pipe are synchronised safely.
