Re: Smoke [ebcdic] v5.21.9-301-g1af0e69 FAIL(X) os/390 23.00 (2817/)

On 03/18/2015 05:51 AM, Yaroslav Kuzmin wrote:
> Smoke log available https://drive.google.com/file/d/0B5PTttxwo7qAYUNKMXFFQmtlemM
>
> Automated smoke report for branch ebcdic 5.21.10 patch 1af0e6947339da7d49c01af0b369c0b8641484e8 v5.21.9-301-g1af0e69
> RS12: 2817 (2817/)
>      on        os/390 - 23.00
>      using     c99 version
>      smoketime 6 hours 50 minutes (average 3 hours 25 minutes)
>
> Summary: FAIL(X)
>
> O = OK  F = Failure(s), extended report at the bottom
> X = Failure(s) under TEST but not under harness
> ? = still running or test results not (yet) available
> Build failures during:       - = unknown or N/A
> c = Configure, m = make, M = make (after miniperl), t = make test-prep
>
> v5.21.9-301-g1af0e69  Configuration (common) none
> ----------- ---------------------------------------------------------
> X X X X
> | | | +----- PERLIO = perlio -DDEBUGGING
> | | +------- PERLIO = stdio  -DDEBUGGING
> | +--------- PERLIO = perlio
> +----------- PERLIO = stdio
>
>
> Locally applied patches:
>      SMOKE1af0e6947339da7d49c01af0b369c0b8641484e8
>
> Tests skipped on user request:
>      # One test name on a line
> Failures: (common-args) none
> [stdio]
> Inconsistent test results (between TEST and harness):
>      ../t/re/pat_re_eval.t................... .............................................. FAILED--expected 527 tests, saw 297
>      ../lib/meta_notation.t.................. ............................................. FAILED at test 2
>      ../t/op/pack.t.......................... ..................................................... FAILED--expected 14708 tests, saw 14709
>
> [perlio]
> [stdio] -DDEBUGGING
> Inconsistent test results (between TEST and harness):
>      ../lib/meta_notation.t.................. ............................................. FAILED at test 2
>      ../t/op/pack.t.......................... ..................................................... FAILED--expected 14708 tests, saw 14709
>      ../t/re/pat_re_eval.t................... .............................................. FAILED--expected 527 tests, saw 297
>
> [perlio] -DDEBUGGING
> Inconsistent test results (between TEST and harness):
>      ../t/op/pack.t.......................... ..................................................... FAILED--expected 14708 tests, saw 14709
>      ../lib/meta_notation.t.................. ............................................. FAILED at test 2
>      ../t/re/pat_re_eval.t................... .............................................. FAILED--expected 527 tests, saw 297
>
> Compiler messages(os390):
>
>
>
> --
> Report by Test::Smoke v1.6 running on perl 5.21.7
> (Reporter v0.052 / Smoker v0.045)
>
>

This report shows that changing the name of mro.c worked.  I plan to 
push that to blead tomorrow if no one objects, and it doesn't look like 
anyone will.

I have pushed a new branch to test.  This will be the final possible 
test before the code freeze.

The pack.t problem was a clerical error, fixed in the new branch.

The meta_notation problem is very puzzling.  I've fixed one error, but 
it looks like the other one is from some underlying bug that just now is 
showing up.  I've added debugging info in the new branch.  If worse 
comes to worse, these tests can be marked as TODO.

The pat_re_eval.t failure appears to be related to Encode.  We're not 
fixing Encode to work very well in v5.22, and so in yesterday's branch I 
removed the partial fixes that have been in it for quite some time, in 
order to see what problems might show up.  And this is the only one.  So 
the new branch just skips those tests.

The other known problem that will be in 5.22 and hasn't been discussed 
recently is that Configure has this line
#! /bin/sh
and on your system /bin/sh is a ksh that doesn't work very well with 
Configure.  For your convenience, I've had a patch to change that line 
to point to a bash (and that remains in the current branch to test), but 
it won't get into blead.  When doing things manually, one can just call 
Configure with a shell it works well with, but I believe Test::Smoke 
doesn't work well without this fix.  That is really outside the scope of 
the Perl core, except we need to make sure that smoking happens, so 
we'll work something out after the freeze.  I too use a ksh, but mine 
works fine with Configure.

Is it typical for /bin/sh to be a ksh on z/OS?