Folks:

The current unstable (darcs-2-pre) fails its self-tests for me.  The  
last part of the output is:

Test Summary Report
-------------------
pull.pl                   (Wstat: 65280 Tests: 14 Failed: 0)
   Non-zero exit status: 255
   Parse errors: No plan found in TAP output
Files=39, Tests=1536, 103 wallclock secs ( 0.48 usr  0.20 sys + 21.77  
cusr 39.13 csys = 61.58 CPU)
Result: FAIL
make: *** [test_perl] Error 1

The output from make test contains the amusing snippet:

push..........................ok
pull_many_files...............ok
pull..........................Invalid value for shared scalar at / 
System/Library/Perl/5.8.6/Test/Builder.pm line 319.
WHOA!  Somehow you got a different number of results than tests ran!
This should never happen!  Please contact the author immediately!
END failed--call queue aborted.
  Dubious, test returned 255 (wstat 65280, 0xff00)
  All 14 subtests passed
printer.......................ok
pending_has_conflicts.........ok

Attached are full logs.

Regards,

Zooko

Attachments

• testfaillog.7z

Thanks for the report, Zooko. 

I'll take a first whack at this from a Perl perspective. Looks like it might be
a test-suite problem, and not a real issue.

Zooko,

I'm sending two patches to darcs-devel and cc'ing it to you. I suspect this
patch will fix the issue, but it is a "good measure" anyway. 

One patches causes us to always prefer our internal perl libraries, which will
help give more consistent results.  The second patch freshens up some of the
Perl testing libraries for good measure.

   Mark

Mark, with your patches for this bug, I suddenly *get* the bug on my system!

In particular, I'm running the tests with:

  *  issue602: part 1: Always prefer our private copy of Test::More over the
system-wide one for more consistent results

Tue Jan 22 20:36:42 EST 2008  Mark Stosberg <mark@summersault.com>
  *  issue602, part 2: freshen our versions of Test::More and Test::Builder

and I get

pending_has_conflicts..................Undefined subroutine &main::like called
at pending_has_conflicts. pl line 22.
# Looks like your test died before it could output anything.
pending_has_conflicts.................. Dubious, test returned 255 (wstat 65280,
0xff00)
 No subtests run


and

Test Summary Report
-------------------
pending_has_conflicts.pl           (Wstat: 65280 Tests: 0 Failed: 0)
  Non-zero exit status: 255
  Parse errors: No plan found in TAP output
Files=40, Tests=1515, 87 wallclock secs ( 0.60 usr  0.07 sys + 33.01 cusr 20.47
csys = 54.15 CPU)
Result: FAIL
make: *** [test_perl] Error 123

I should note that without these two patches, the test suite works just fine.

David Roundy wrote:
> 
> Mark, with your patches for this bug, I suddenly *get* the bug on my system!
> 
> pending_has_conflicts..................Undefined subroutine &main::like called
> at pending_has_conflicts. pl line 22.
> # Looks like your test died before it could output anything.
> pending_has_conflicts.................. Dubious, test returned 255 (wstat 65280,

I think this was a mistake in the what I sent. 
"pending_has_conflicts.pl" had some in-progress updates on my system 
that I did not mean to send. You could roll-back the changes to that 
file, or I could amend-record.

This is actually good news. It means all other tests including 
previously failing "match.pl" passed for you.

   Mark

On Wed, Jan 23, 2008 at 03:32:22PM -0000, Mark Stosberg wrote:
> David Roundy wrote:
> > Mark, with your patches for this bug, I suddenly *get* the bug on my system!
> > 
> > pending_has_conflicts..................Undefined subroutine &main::like called
> > at pending_has_conflicts. pl line 22.
> > # Looks like your test died before it could output anything.
> > pending_has_conflicts.................. Dubious, test returned 255 (wstat 65280,
> 
> I think this was a mistake in the what I sent. 
> "pending_has_conflicts.pl" had some in-progress updates on my system 
> that I did not mean to send. You could roll-back the changes to that 
> file, or I could amend-record.

If you could send an amend-recorded version, that would be great.

> This is actually good news. It means all other tests including 
> previously failing "match.pl" passed for you.

Actually, match.pl has been passing for me (at least during the day while
I'm at work...).  But yeah, it does seem good.
-- 
David Roundy
Department of Physics
Oregon State University

I have sent an amend-recorded patch now to darcs-devel.

Zooko, could you test with the new "issue602" patches?

This is believed to be fixed now, but a confirmation from an OS X/intel user
would be helpful.