docker exec vs cmd modes

```
            | No ENTRYPOINT   | ENTRYPOINT entry                 | ENTRYPOINT ["entry"]  |         
------------+-----------------+------------------------------ ---+-----------------------|
No CMD      | <error>         | /bin/sh -c entry                 | entry                 |
CMD cmd     | /bin/sh -c cmd  | /bin/sh -c entry /bin/sh -c cmd  | entry /bin/sh -c cmd  |
CMD ["cmd"] | cmd             | /bin/sh -c entry cmd             | entry cmd             |
------------------------------------------------------------------------------------------
```

An additional dimension is missing:
Are command-line args provided to "docker run CONTAINER" or not (and should they be used or ignored).
In this case only both exec modes (lower right cell) works as expected for me.

Therefore, there are only 2 use cases (IMHO, independent of the combinations above):

• container supports command-line args (and command-line args may override default args)
  => USE: Both exec mode.
• container should ignore any command-line args in "docker run CONTAINER ..."
  => USE: Other cases (but use either only CMD or only ENTRYPOINT for readability/clarity).
  BETTER: Use only ENTRYPOINT in shell mode (other cases lead to weird problems).