test suites, again

[Zack Weinberg]

I've put another version of my new test suite up for anonymous FTP.

rabi.phys.columbia.edu:/pub/zack/testsuite-0.3.tar.gz

This one is complete: all but eight of the test programs from the old
testsuite are run.  Of those eight, one is Mach-specific, six are
programming examples not proper test cases, and one (xbug.c) I have no
idea what it's testing.  I don't think we lose terribly much by
turning them off (of course if someone wants to figure them out and
fit them into the new framework that'd be great).

The .diff file in the same directory adds some necessary wrapper
headers.  Ulrich put this into CVS, but it's there just in
case. There's a doc file in the tarball (Texinfo source, this time).

In several cases - notably printf, getopt, and argp - the test
framework was constructed by assuming that the present behavior is
correct.  This may not be true.  I'd appreciate it if people would
check over the tests to verify that they are correct.  The "expected
failures" indicate behavior I'm certain is wrong.

When run, make will generate a bunch of files and then invoke runtest,
which should produce output similar to the appended example.  If you
get anything substantially different please tell me.

I'm going off-net until September.  Have a good month all.

zw

Test run by zack on Sat Aug 01 01:35:11 1998
Test of libc 2.0.95 for i586-pc-linux-gnu
    compiled by gcc version egcs-2.90.29 980515 (egcs-1.0.3 release)

Running coverage tests...

Running regression tests...

Running special tests...

Running standards tests...

Summary:
# of expected passes            983
# of expected failures          39
# of unsupported testcases      12

Test completed at Sat Aug 01 01:52:42 1998
Time elapsed: 17m 31s