I have build version of Pearl (git commit 2877e8d ) , where i made a rough 
transfer of all functions in the file ../ext/mro/mro.c (I know that this file 
is generated from the file ../ext/mro/mro.xs) in file ../mro.c.
I completely deleted mro module and transferred all its functionality into the 
main library libperl.a.
In attached files log commands 'make test' and two modified files module mro.
Number of passed tests increased to 94%.

With the passage of the main test (make test) for passing the test 
t/re/fold_grind.t
Perl asked for too much memory and the system stopped working with the main 
test report 'Out of memory!'  but separate test (t/re/fold_grind.t)  passage. 
When removed test t/re/fold_grind.t from the catalog all the other tests passed.


P.S. Last week I sent a email to Peter, but he has not written anything in the 
mailing list.

------------------------------------------------------------------------
 Yaroslav Kuzmin
Developer C/C++ ,z/OS , Linux
3 Zhukovskiy Street · Miass, Chelyabinsk region 456318 · Russia
Tel:  +7.922.2.38.33.38
Email: [email protected]
Web: www.rocketsoftware.com

________________________________________
От: Karl Williamson <[email protected]>
Отправлено: 19 августа 2014 г. 1:24
Кому: Yaroslav Kuzmin; Perl5 Porters; perl-mvs
Тема: Re: Bug Report z/OS

On 08/15/2014 03:09 AM, Yaroslav Kuzmin wrote:
>
> git commit 09861a101b2aabfee2ad22840d97fd16e04c03f0
>       Author: Karl Williamson <[email protected]>
>       Date:   Tue Jul 29 20:47:15 2014 -0600
>
> Make test
>
> ------------------------------------------------------------------------
>   Yaroslav Kuzmin

Note that this report has 8000 fewer error lines than the previous one.
  I consider that good progress. And for the first time ever, /i regex
matching works on EBCDIC.

To continue progressing:

1) There is still a problem with you somehow getting old stuff.  The
make output you attached includes these lines:
t/lib/charnames/alias .................................................
PROG:
# Below, one of the EXPECT regexes matches both the UTF-8 and non-UTF-8
form.
# This is because under some circumstances the message gets output as UTF-8.
# We use \xab, as that is invalid in both ASCII and EBCDIC platforms, and we
# accept both UTF-8 and 1047 UTF-EBCDIC
use charnames ":full", ":alias" => {
                             "4e_ACUTE" => "LATIN SMALL LETTER E WITH
ACUTE",
                             "e_A,CUTE" => "LATIN SMALL LETTER E WITH
ACUTE",
                             "e_ACUT\x{ab}E" => "LATIN SMALL LETTER E
WITH ACUTE",
                     };

But the commit you said you tested above
09861a101b2aabfee2ad22840d97fd16e04c03f0 changes those lines, thusly:
-# NAME various wrong characters in :alias are errors
-# Below, one of the EXPECT regexes matches both the UTF-8 and non-UTF-8
form.
-# This is because under some circumstances the message gets output as
UTF-8.
-# We use \xab, as that is invalid in both ASCII and EBCDIC platforms,
and we
-# accept both UTF-8 and 1047 UTF-EBCDIC
+# NAME A wrong character in :alias is an error
+# These next tests could be combined, but the messages can come out in
+# different orders on EBCDIC vs ASCII, and can't have both 'random' and
'regex'
+# options, and need 'regex' to avoid 'at line X' getting in the way.
  use charnames ":full", ":alias" => {
                              "4e_ACUTE" => "LATIN SMALL LETTER E WITH
ACUTE",
-                            "e_A,CUTE" => "LATIN SMALL LETTER E WITH
ACUTE",
-                            "e_ACUT\x{ab}E" => "LATIN SMALL LETTER E
WITH ACUTE",
                      };

Thus you were somehow using an old version of t/lib/charnames/alias.  I
can't figure out how that could happen.  Other things in the output
indicate you were testing things that were included in this single commit.

2) I think we would be well above the 90% pass rate if the mro stuff
would work. Nicholas wrote some suggestions about how to get it to, in
<[email protected]>
Did you try this?

2) I know that there are bugs in pack/unpack, which I have deferred
fixing until other things settle down.  I think we are almost there!
Various remaining test failures should go away when that gets fixed.  As
a start to looking at this, I need the current contents of your
directory lib/unicore.  Just email it in a convenient format for you to
generate.

3) It would be helpful to have you using a DEBUGGING build.  Add
-DDEBUGGING to Configure

4) I have made a number of fixes to this branch, and pushed a new one to
the same place
http://perl5.git.perl.org/perl.git/shortlog/refs/heads/khw/ebcdic
with the topmost commit being
http://perl5.git.perl.org/perl.git/commit/2877e8d28f6866ecd9c385e7205327bb55420a48

Some of these fixes fix tests, and some fix code bugs that should cause
tests to now pass that previously didn't.  Some add extra tests to help
me figure out what's going wrong.




================================
Rocket Software, Inc. and subsidiaries ■ 77 Fourth Avenue, Waltham MA 02451 ■ 
+1 800.966.3270 ■ +1 781.577.4321
Unsubscribe From Commercial Email – [email protected]
Manage Your Subscription Preferences - 
http://info.rocketsoftware.com/GlobalSubscriptionManagementEmailFooter_SubscriptionCenter.html
Privacy Policy - http://www.rocketsoftware.com/company/legal/privacy-policy
================================

Attachment: make_test.log.tar.gz
Description: make_test.log.tar.gz

Attachment: mro.tar.gz
Description: mro.tar.gz

Reply via email to