@hackage shelltestrunner0.4

Run a given program through "shell" tests specifed by one or more test files, each of which can specify: command-line arguments, input, expected output, expected stderr output, and expected exit code. This was extracted from the hledger project, inspired by the tests in John Wiegley's ledger project, and uses test-framework's test runner.

Usage:

shelltestrunner [opts] executable testfile1 [testfile2 ...] [-- <test-framework runner opts>]

You can pass options through to test-framework's runner; they must go after -- at the end. You may be able to get a big speedup by running tests in parallel: try -- -j8.

Test file format:

# 0 or more comment lines beginning with #
-opt1 -opt2 arg1 arg2 # command line args, executable will be prepended
<<<
0 or more lines of input
>>> [/regexp/]
[..or 0 or more lines of expected output]
>>>2 [/regexp/]
[..or 0 or more lines of expected error output]
>>>= [/regexp/]
[..or expected numeric exit code]

The expected fields can also have a regular expression match expression following the delimiter on the same line, with no other data lines, in which case the test passes if the output is matched by the regexp. The regexp is enclosed in forward slashess. A ! preceding the expression negates the match. For example, to check that stdout does not contain "axe":

>>> !/axe/

Apart from the command line, all fields are optional. Only fields specified in the test will be tested, unless you use the -i/--implicit-tests flag, which will test for empty stdout, empty stderr, or 0 exit code whenever fields are omitted.

Issues:

• output order is mixed up

• option processing is weak

• can't test input/output which does not end with newline

Wishlist:

• configurable delimiters

• multiple tests per file