[Monotone-devel] lua testsuite progress (was: Re: Summer of Code kick-off)

[Timothy Brownawell]

Testsuite progress is going much faster than I'd expected, I've
currently got 200 of the 346 test converted.

It now works under Windows/VC8, except for the non-english tests. Both
of these fail with "failed to convert string from iso-8859-1 to UTF-8:
'<o-umlat><a-umlat><u-umlat><double-s-thing-that-looks-like-B>'". (The
input string is "\xF6\xE4\xFC\xDF", or "\246\228\252\223".) It works
under Linux, so I'm currently assuming it's something wrong with
VC8-compiled monotone.

Next up is to convert more tests (maybe all of the rest?) and get the
*nix-ish build env installed on my windows so I can verify that things
work there. I can also (but probably not this week) poke at our
dependency graph more, to try to get the test program to depend less on
general monotone code (besides sanity and libplatform) (maybe enough to
have a 'libsanity'?).

The 200 converted tests run in 27 minutes walltime on Windows/vmplayer,
with 10 skipped (These are either tests that don't make sense on
Windows, or tests that use unavailable commands (such as the cvs
tests).) On Linux, they run in 9 minutes walltime, with 5 minutes user
and 1 minutes system time. The first 200 tests from the old testsuite
(on Linux) take 20m/5m/10m wall/user/sys time.

Tim


=====

The empty_environment test probably won't work on Windows (but it's
skipped if you don't have a *nix-ish userland). So it has a check(false)
at the end, until I fix it for that.

=====

One test, t_rcs_import.at (now test_problematic_cvs_import), was broken
(because a ,v filename wasn't quoted where it should have been). So that
test has gone from pass to xfail.

-- This test imports a rcs file which fails according to the mailing
list
-- reporter: Carl Christian Kanne "Bug? in CVS import monotone 0.18"
-- Date:        Fri, 15 Apr 2005 12:53:13 +0200

The failed import says:
mtn: ticks: b="branches"/1, v="versions"/1
mtn: bvbvvvvv
mtn: warning: error reading RCS file rcsfile,v: out_of_range while
processing a102 58
mtn: warning:  with source.size() == 101 and cursor == 102
mtn: ticks: b="branches"/1, r="revisions"/1, v="versions"/1
mtn: rrrrrrrtt

=====

Also, the merge_5 test has gone from xfail to pass, since we now
generate a conflict instead of a mismerge. Apparently there actually
isn't a single correct way to merge it.

http://colabti.de/irclogger/irclogger_log/monotone?date=2006-06-07,Wed#l98