This patch improves the SSH multi line splitter parser to take into some
cases that were not covered as well as fix some of the tests.

Code fixes:

- On slow SSH connections we may not get the initial dump of stdin in
  the response, so looking for the "exit" command we sent will never
  find it and the parsing will fail.  We now look for the prompt and
  ignore whether we have the stdin data or not, since it's not really
  relevant.

- Sometimes we get empty prompts in the SSH response, which either break
  the current parsing or makes us return bad data.  We now remove all
  empty prompts.

Code improvements:

- We were always removing the last 2 lines from the response without
  actually checking if these 2 lines are the exit command and an empty
  line.  Now we search for the final exit command that we execute to
  determine the end of the valid output data.

- If the output data has any "\r" at the end of a line the parser
  returns them, which we shouldn't. They should be removed just like we
  remove the "\r\n" when splitting the lines.

Test fixes in HPE3ParClientMockSSHTestCase::test_strip_input_from_output:

- In a couple of places tests are passing [cmd, X, Y] to the parser
  expecting a failure, which we get, and we think everything is working
  as expected, but the failure is not for the reasons we want, so we are
  not testing anything.  By passing that data we are effectively passing
  [['foo', '-v'], X, Y] which will break the parser because that's
  "garbage".

- The data passed and expected in the success case can be misleading, as
  it doesn't include the exit prompt and doesn't expect the footer to be
  returned, but the footer should be returned and parsing a command that
  doesn't have the last exit command is not valid.

Closes hpe-storage#78

This patch improves the SSH multi line splitter parser to take into
account some cases that were not covered as well as fix some of the tests.

Code fixes:

- On slow SSH connections we may not get the initial dump of stdin in
  the response, so looking for the "exit" command we sent will never
  find it and the parsing will fail.  We now look for the prompt and
  ignore whether we have the stdin data or not, since it's not really
  relevant.

- Sometimes we get empty prompts in the SSH response, which either break
  the current parsing or makes us return bad data.  We now remove all
  empty prompts.

Code improvements:

- We were always removing the last 2 lines from the response without
  actually checking if these 2 lines are the exit command and an empty
  line.  Now we search for the final exit command that we execute to
  determine the end of the valid output data.

- If the output data has any "\r" at the end of a line the parser
  returns them, which we shouldn't. They should be removed just like we
  remove the "\r\n" when splitting the lines.

Test fixes in HPE3ParClientMockSSHTestCase::test_strip_input_from_output:

- In a couple of places tests are passing [cmd, X, Y] to the parser
  expecting a failure, which we get, and we think everything is working
  as expected, but the failure is not for the reasons we want, so we are
  not testing anything.  By passing that data we are effectively passing
  [['foo', '-v'], X, Y] which will break the parser because that's
  "garbage".

- The data passed and expected in the success case can be misleading, as
  it doesn't include the exit prompt and doesn't expect the footer to be
  returned, but the footer should be returned and parsing a command that
  doesn't have the last exit command is not valid.

Closes hpe-storage#78

This patch improves the SSH multi line splitter parser to take into
account some cases that were not covered as well as fix some of the tests.

Code fixes:

- On slow SSH connections we may not get the initial dump of stdin in
  the response, so looking for the "exit" command we sent will never
  find it and the parsing will fail.  We now look for the prompt and
  ignore whether we have the stdin data or not, since it's not really
  relevant.

- Sometimes we get empty prompts in the SSH response, which either break
  the current parsing or makes us return bad data.  We now remove all
  empty prompts.

Code improvements:

- We were always removing the last 2 lines from the response without
  actually checking if these 2 lines are the exit command and an empty
  line.  Now we search for the final exit command that we execute to
  determine the end of the valid output data.

- If the output data has any "\r" at the end of a line the parser
  returns them, which we shouldn't. They should be removed just like we
  remove the "\r\n" when splitting the lines.

Test fixes in HPE3ParClientMockSSHTestCase::test_strip_input_from_output:

- In a couple of places tests are passing [cmd, X, Y] to the parser
  expecting a failure, which we get, and we think everything is working
  as expected, but the failure is not for the reasons we want, so we are
  not testing anything.  By passing that data we are effectively passing
  [['foo', '-v'], X, Y] which will break the parser because that's
  "garbage".

- The data passed and expected in the success case can be misleading, as
  it doesn't include the exit prompt and doesn't expect the footer to be
  returned, but the footer should be returned and parsing a command that
  doesn't have the last exit command is not valid.

Closes hpe-storage#78