Simple NFT MONITOR Testsuite

============================

The purpose of this suite of tests is to assert correct 'nft monitor' output for

known input. The suite consists of the single shell script 'run-tests.sh' which

performs the tests and a number of test definition files in 'testcases/'. The

latter have to be suffixed '.t' in order to be recognized as such.

Test Case Syntax

----------------

Each testcase defines a number of commands to pass on to 'nft' binary and an

associated 'nft monitor' output definition. Prerequisites for each command have

to be established manually, i.e. in order to test monitor output when adding a

chain, the table containing it has to be created first. In between each

testcase, rule set is flushed completely.

Input lines are prefixed by 'I'. Multiple consecutive input lines are passed to

'nft' together, hence lead to a single transaction.

There are two types of output lines: Those for standard syntax, prefixed by 'O'

and those for JSON output, prefixed by 'J'. For standard syntax output lines,

there is a shortcut: If a line consists of 'O -' only, the test script uses all

previous input lines as expected output directly. Of course this is not

available for JSON output lines.

Empty lines and those starting with '#' are ignored.

Test Script Semantics

---------------------

The script iterates over all test case files, reading them line by line. It

assumes that sections of 'I' lines alternate with sections of 'O'/'J' lines.

After stripping the prefix, each line is appended to a temporary file. There are

separate files for input and output lines.

If a set of input and output lines is complete (i.e. upon encountering either a

new input line or end of file), a testrun is performed: 'nft monitor' is run in

background, redirecting the output into a third file. The input file is passed

to 'nft -f'. Finally 'nft monitor' is killed and it's output compared to the

output file created earlier. If the files differ, a unified diff is printed and

test execution aborts.

After each testrun, input and output files are cleared.

Note: Running 'nft monitor' in background is prone to race conditions. Hence

an artificial delay is introduced before calling 'nft -f' to allow for 'nft

monitor' to complete initialization and another one before comparing the output

to allow for 'nft monitor' to process the netlink events.

By default, only standard syntax is being tested for, i.e. 'J'-prefixed lines

are simply ignored. If JSON testing was requested (by passing '-j' flag to the

test script), 'O'-prefixed lines in turn are ignored.

There is one caveat with regards to JSON output: Since it always contains handle

properties (if the given object possesses such) which is supposed to be

arbitrary, there is a filter script which normalizes all handle values in

monitor output to zero before comparison. Therefore expected output must have

all handle properties present but with a value of zero.