| Subject: | kadmin should have a script-friendly mode |
The following design should be relatively easy to implement and would
make kadmin and kadmin.local easier to use in administrative scripts,
without creating backwards compatibility issues.
If there are any command-line arguments after the kadmin_startup getopt
loop is complete, they should be interpreted as a command with the
following semantics:
1. The command is split according to the argv entries (i.e. using
ss_execute_command), not according to ss_execute_line quoting.
2. No informational messages are displayed to stdout. Error messages
can still be displayed to stderr.
3. If the command fails to complete, kadmin exits with a non-zero
status. (In normal use, even with -q, it only exits with nonzero status
if it fails to connect to kadmind.)
I don't propose to fix #743 (prompt is displayed to stdout). I expect
this to be most useful in combination with kadmin -c so that the script
doesn't have to deal with inputting a password on each kadmin
invocation.
Converting our Python tests to use the new syntax would be a bonus, as
it would mean the tests would immediately notice when a kadmin command
fails.
make kadmin and kadmin.local easier to use in administrative scripts,
without creating backwards compatibility issues.
If there are any command-line arguments after the kadmin_startup getopt
loop is complete, they should be interpreted as a command with the
following semantics:
1. The command is split according to the argv entries (i.e. using
ss_execute_command), not according to ss_execute_line quoting.
2. No informational messages are displayed to stdout. Error messages
can still be displayed to stderr.
3. If the command fails to complete, kadmin exits with a non-zero
status. (In normal use, even with -q, it only exits with nonzero status
if it fails to connect to kadmind.)
I don't propose to fix #743 (prompt is displayed to stdout). I expect
this to be most useful in combination with kadmin -c so that the script
doesn't have to deal with inputting a password on each kadmin
invocation.
Converting our Python tests to use the new syntax would be a bonus, as
it would mean the tests would immediately notice when a kadmin command
fails.