…ut content, dependencies, and UPI (unique patch ID) differ. These patches have to be rolled back in advance of applying the PSU/RU and must afterwards be installed with the correct version (=UPI). They should however not be rolled back if they are already there with the desired UPI. This helps to write more idempotent playbooks.

So there is a new parameter exclude_upi that is evaluated only if state=absent.

To keep changes in working code at a minimum the check for UPI could be acquired only by another "opatch lspatches" that costs some seconds.

Oracle_opatch searches the running listener with:

```ps -elf | grep "[0-9] $ORACLE_HOME/bin/tnslsnr"```

...then splits the line by spaces and takes the last but one field.
This does not work for a "ps -ef" result like this:

```0 S oracle   16921     1  0  80   0 - 43851 ep_pol 21:58 ?        00:00:00 /u01/app/oracle/product/12.1.0.2/ee/bin/tnslsnr LISTENER /dev/null -inherit```

In this case "linelist[-2]" deliveres "/dev/null" instead of the LISTENER
name. My fix is to search the ps line for the listener executable and use
the next field.

Signed-off-by: Dietmar Uhlig <[email protected]>

Signed-off-by: Dietmar Uhlig <[email protected]>

(I should have tested more thoroughly after the last refactoring.)

Merging Mikael Sandström's current version into my patched version.

…acle-modules

…ut content, dependencies, and UPI (unique patch ID) differ. These patches have to be rolled back in advance of applying the PSU/RU and must afterwards be installed with the correct version (=UPI). They should however not be rolled back if they are already there with the desired UPI. This helps to write more idempotent playbooks.

So there is a new parameter exclude_upi that is evaluated only if state=absent.

To keep changes in working code at a minimum the check for UPI could be acquired only by another "opatch lspatches" that costs some seconds.

…AL_LOG_DATA_MIN

…ules