vtable_massacre branch needs more eyes

Andrew Whitworth wknight8111 at gmail.com
Sat Feb 13 15:46:36 UTC 2010


I'll take a look at those tests in more detail to find out where the
platform-specific differences in the tests are happening.

--Andrew Whitworth



On Sat, Feb 13, 2010 at 9:01 AM, James E Keenan <jkeen at verizon.net> wrote:
> Andy Dougherty wrote:
>>
>> With this branch at Revision: 43921 on Solaris, I got two new sets of
>> failure.  Both look like bad plans -- the "List of Failed" are all greater
>> than the "Total".
>>
>> Failed Test            Stat Wstat Total Fail  List of Failed
>>
>> -------------------------------------------------------------------------------
>> t/op/arithmetics_pmc.t               44   12  45-56
>> t/pmc/bigint.t                       34   11  35-45
>
> r43921 was the point at which the branch was merged back into trunk.  At
> this same revision, I am getting the same problem as Andy in t/pmc/bigint.t
> on Darwin/PPC, but not on Linux/i386.  (I am not getting failures in
> t/op/arithmetics_pmc.t on either box.)
>
> This is the output I'm getting on Darwin/PPC:
>
> $ prove -v t/pmc/bigint.t
> t/pmc/bigint.t ..
> 1..34
> This Parrot uses GMP
> Buggy GMP version [4.1.3] with huge digit multiply - please upgrade
> ok 1 #skip skipped
> ok 2 #skip skipped
> ok 3 #skip skipped
> ...
> ok 44 #skip skipped
> ok 45 #skip skipped
> All 34 subtests passed
>        (less 45 skipped subtests: -11 okay)
>
> Test Summary Report
> -------------------
> t/pmc/bigint.t (Wstat: 0 Tests: 45 Failed: 11)
>  Failed tests:  35-45
>  Parse errors: Bad plan.  You planned 34 tests but ran 45.
> Files=1, Tests=45,  4 wallclock secs ( 0.13 usr  0.05 sys +  0.12 cusr 0.11
> csys =  0.41 CPU)
> Result: FAIL
>
> Now, I've been getting that "Buggy GMP" message -- and hence, SKIPping tests
> -- on this test file for years.  But I never got a FAIL due to some problem
> in test counting until now.  When this was still in the branch, I got this
> problem on Darwin/PPC and changed the number of tests in the plan to 45, at
> which point all the tests were correctly SKIPped and the file as a whole
> PASSed.
>
> But apparently that was the wrong thing to do and someone changed it back to
> 34 before the merge.  (I can't say who, because this is one of those files
> on which 'svn blame' refuses to work, saying that the file is "binary."  I
> was told once upon a time that this was due to a defective CVS-to-SVN
> conversion way back when.)
>
> kid51
>
> _______________________________________________
> http://lists.parrot.org/mailman/listinfo/parrot-dev
>


More information about the parrot-dev mailing list