通过curl登录在bash脚本中失败，相同的curl在命令行中成功

Question

I'm running this login via curl in my bash script. I want to make sure I can login before executing the rest of the script, where I actually log in and store the cookie in a cookie jar and then execute another curl in the API thousands of times. I don't want to run all that if I've failed to login.

Problem is, the basic login returns 401 when it runs inside the script. But when I run the exact same curl command on the command line, it returns 200!

basic_login_curl="curl -w %{http_code} -s -o /dev/null -X POST -d \"username=$username&password=$password\" $endpoint/login"
echo $basic_login_curl
outcome=`$basic_login_curl`
echo $outcome
if [ "$outcome" == "401" ]; then
    echo "Failed login. Please try again."; exit 1;
fi

This outputs:

curl -w %{http_code} -s -o /dev/null -X POST -d "username=bdunn&password=xxxxxx" http://stage.mysite.it:9301/login
401
Failed login. Please try again.

Copied the output and ran it on the cmd line:

$ curl -w %{http_code} -s -o /dev/null -X POST -d "username=bdunn&password=xxxxxx" http://stage.mysite.it:9301/login
200$

Any ideas? LMK if there's more from the code you need to see.

ETA: Please note: The issue's not that it doesn't match 401, it's that running the same curl login command inside the script fails to authenticate, whereas it succeeds when I run it on the actual CL.

Answer 1

Most of the issues reside in how you are quoting/not quoting variables and the subshell execution. Setting up your command like the following is what I would recommend:

basic_login_curl=$(curl -w "%{http_code}" -s -o /dev/null -X POST -d "username=$username&password=$password" "$endpoint/login")

The rest basically involves quoting everything properly:

basic_login_curl=$(curl -w "%{http_code}" -s -o /dev/null -X POST -d "username=$username&password=$password" "$endpoint/login")
# echo "$basic_login_curl" # not needed since what follows repeats it.
outcome="$basic_login_curl"
echo "$outcome"
if [ "$outcome" = "401" ]; then
    echo "Failed login. Please try again."; exit 1;
fi

Running the script through shellcheck.net can be helpful in resolving issues like this as well.