Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] View of /tests/trunk/README
ViewVC logotype

View of /tests/trunk/README

Parent Directory Parent Directory | Revision Log Revision Log

Revision 2071 - (download) (annotate)
Tue Oct 31 02:58:12 2006 UTC (12 years, 7 months ago) by blume
File size: 4725 byte(s)
moved tests trunk
		   SML/NJ Regression Testing

This directoy contains regression test suites and shell scripts for
running them.

There are 5 active test suites, found in the directories:

  coresml     core SML tests (based on tests from Peter Sestoft)
  typing      type checking
  modules     modules
  bugs	      bug tests
  basis	      Basis tests (based on tests from Peter Sestoft) [Note: filesys]

(the sixth directory, "printing", is not currently active).

Each test suite directory contains two subdirectories

  tests	      source files for tests
  outputs     corresponding output files [Note: output variants]

and a file


that lists test files with known problems in the current version
of SML/NJ (i.e. known regression failures).  [Note!: basis/tests setup]


  bin/testml.sh		- performs tests in a suite
  bin/process.sh	- generates report of results
  bin/dotest.sh		- combines testml.sh and process.sh
  bin/doallseq.sh 	- performs all suites one after another
  bin/doallpar.sh	- performs all suites in parallel

Normally one will execute one of dotest.sh, doallseq.sh. or
doallpar.sh.  The scripts are designed not to interfere with one
another, so that test suites can be executed concurrently on a given
architecture, and also on multiple machines (or varying architectures)
sharing the file system containing this directory.

For a particular architecture and operating system, arch-os,
bin/testml.sh produces a file <testdir>/LOG.arch-os and
a directory <testdir>/bad.arch-os.  process.sh <testdir> then
produces a file <testdir>/RESULTS.arch-os.  

Before running a regression test suite, any old copies of these files
and directories should be removed (if not, the scripts will complain).

bin/dotest.sh <testdir> [-diff] [-sml <filename>]  -  run tests in <testdir>
    the name of the test suite to be run (coresml, typing, etc).

  -sml <filename>:
    the sml command to be used (e.g. "/usr/local/sml/bin/sml").
    If this option is not supplied, the value of shell variable SML
    is used instead, and if that is not defined, the default command
    name "/usr/local/sml/bin/sml" is used.

    this option causes the diff between the test output and 
    the corresponding reference output (in <testdir>/outputs) to be
    used in the RESULT file, overriding the default behavior, which
    is to include the old and new outputs in full.

bin/doallseq.sh [-sml <filename>]	  - run all suites sequentially
bin/doallpar.sh [-sml <filename>]	  - run all suites in parallel

  -sml <filename>:
    the sml command to be used (e.g. "/usr/local/sml/bin/sml").
    If this option is not supplied, the value of shell variable SML
    is used instead, and if that is not defined, the default command
    name "/usr/local/sml/bin/sml" is used.

Example: (with the testing directory as current working directory)

  rm -rf bugs/LOG.sparc-solaris bugs/bad.sparc-solaris bugs/RESULTS.sparc-solaris
  (or just rm -rf bugs/*.sparc-solaris)

  bin/dotest.sh bugs -diff -sml /usr/local/sml/110/bin/sml

Then review the bugs/RESULTS.sparc-solaris file to see if there are
any discrepancies, which could be unexpected failures or bug fixes.


[filesys] The test basis/tests/filesys.sml is sensitive to file
permissions and may produce spurious failures ("EXN" instead of "OK")
if the tester is not the owner of the files in basis/tests.

[basis/tests setup] After you have created tests directory (say by checking
out from the Yale CVS repository), be sure to read the instructions
in tests/basis/tests/README.  This file explains how to set up some
symbolic and hard links needed by the filesys.sml test.

[output variants] There are actually two versions of the outputs directory
for each test suite.  These are named


The first of these should be used for testing versions before 110.5,
while the second should be used for versions >= 110.5.  Before running
any of the test scripts, output should be defined as a symbolic link
to one of these two directories.  E.g. if testing 110.0.6, in each
test suite directory do:

  rm outputs
  ln -s outputs.pre-110.5 outputs

The reason for the two versions of outputs is that printing changed with
110.5 and there is an extra blank line at the end of each output file.

Authors: Dave MacQueen <dbm@research.bell-labs.com>,
	 Lal George <george@research.bell-labs.com>

ViewVC Help
Powered by ViewVC 1.0.0