source: branches/v4/softs/tests_ccvcache_v4/README @ 995

This is a very basic framework for running regression tests against
a mips32+ccvcache TSAR platform. Here's the layout of this directory:

../../../platforms/tsarv4_mono_mmu
        contains the soclib platform description and simulation executable.
        It should be compiled before starting tests.
        it uses the vci_simhelper component, used to stop the simulation
        with an exit value.  The framework uses this exit status
        (with other things, see below) to determine if a test
        is successfull. The simulator will exit with a non-0 exit status
        after 100000 cycles (and a message to stderr), so the tests are
        expected to be simple.
        The platform has a tty, which is redirected to a file along with
        the simlulator's output. The test framework may search for specific
        strings in this file to determine if a test was successfull.
common/
        Some common functions, ldscripts, etc ... used by the tests.
        It has a subr.S which holds usefull subroutines, such as
        print() and printx().
Makefile.inc
        included from the test's Makefiles. Basic stuff to compile a
        mips32 ELF executable for a test. It expects a test.S,
        add the subr.S from common and eventually additionnal object
        files.
test_*/
        these directories holds the test themselves. They have a run
        script which does what is needed to compile and run a test,
        print "test failed" or "test passed" on stdout, and clean the
        directory if it's successfull. A sample script will call
        make, then run the simulator with srdout/stderr redirected
        to a file. If the simulator exists with a 0 status, this
        script search for some strings in the output (that the
        test program would have printed). This is an additionnal way
        to check that all worked properly.
        Tests are written in assembly; as we're testing for very specific
        conditions (such as an instruction in a branch's delay slot on
        a page boundary), we don't want the compiler to optimise things
        or otherwise reorder instructions.
        If the test was successfull, the 'run' script cleanups the
        directory. Otherwise the 'run.out' file is available to
        start looking at what's wrong.
run_tests
        runs all the test_*/run scripts - the goal being that
        all of them print "test passsed" :). Directories are explicitely
        listed in the run_tests script (order does matter to have a readable
        output)