cpan Data::Dumper is behind blead Data::Dumper #15417

p5pRT · 2016-07-02T22:02:28Z

Migrated from rt.perl.org#128525 (status was 'open')

Searchable as RT128525$

p5pRT · 2016-07-02T22:02:28Z

From @karenetheridge

The latest CPAN release of Data::Dumper is 2.154; in blead it is 2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time: https://rt.cpan.org/Ticket/Display.html?id=12282

p5pRT · 2016-07-06T12:39:53Z

From @ribasushi

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is 2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time: https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

p5pRT · 2016-07-06T12:39:53Z

The RT System itself - Status changed from 'new' to 'open'

p5pRT · 2016-07-08T13:59:03Z

From @ribasushi

On 07/06/2016 02:39 PM, Peter Rabbitson wrote:

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is
2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time:
https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

Still the case almost 5 days after original upload :/

p5pRT · 2016-07-08T14:07:08Z

From mail@steffen-mueller.net

On 07/08/2016 03:58 PM, Peter Rabbitson wrote:

On 07/06/2016 02:39 PM, Peter Rabbitson wrote:

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is
2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time:
https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

Still the case almost 5 days after original upload :/

Sorry, missed that. Been out of order sick for most of that. I'm still
in pretty bad shape now.

Has somebody tried debugging/fixing this?

NB: Normally, I'd say anything as old as 5.12 doesn't require going out
of the way for, but considering virtually everything (usually
inadvertently) depends on DD, this seems worth fixing.

--Steffen

p5pRT · 2016-07-08T14:11:19Z

From @ribasushi

On 07/08/2016 04:06 PM, Steffen Mueller wrote:

Has somebody tried debugging/fixing this?

Not that I am aware (sorry for not doing it myself, I am having my hands
full with another dist).

p5pRT · 2016-07-08T14:38:58Z

From mail@steffen-mueller.net

On 07/08/2016 04:10 PM, Peter Rabbitson wrote:

On 07/08/2016 04:06 PM, Steffen Mueller wrote:

Has somebody tried debugging/fixing this?

Not that I am aware (sorry for not doing it myself, I am having my hands
full with another dist).

Since I'm just about to crash, I'm wondering if I shouldn't just delete
the new version and reindex the old until somebody (me/you/anybody)
steps up to fix this?

BTW, sorry that I skipped the dev release stage this one time... Been a
bit out of it with lots of kid/family and (non-technical) work stuff
going on.

--Steffen

p5pRT · 2016-07-08T15:57:12Z

From @haarg

On Fri Jul 08 07:07:08 2016, mail@steffen-mueller.net wrote:

On 07/08/2016 03:58 PM, Peter Rabbitson wrote:

On 07/06/2016 02:39 PM, Peter Rabbitson wrote:

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is
2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time:
https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

Still the case almost 5 days after original upload :/

Sorry, missed that. Been out of order sick for most of that. I'm still
in pretty bad shape now.

Has somebody tried debugging/fixing this?

NB: Normally, I'd say anything as old as 5.12 doesn't require going out
of the way for, but considering virtually everything (usually
inadvertently) depends on DD, this seems worth fixing.

--Steffen

Bisected to this commit:

commit 31ac59b
Author: Karl Williamson <public@khwilliamson.com>
Date: Wed Mar 13 16:16:14 2013 -0600

Data::Dumper: Generalize for EBCDIC platforms

This extends Data::Dumper to work on EBCDIC platforms. This is just the
code changes. Some .t files will be changed as well, in future commits

This involves some code refactoring especially in the .xs code to
collapse EBCDIC/ASCII handling into one. The if-elsif-elsif-...-else
logic is cleaned up, so that there are fewer branches taken on average.

p5pRT · 2016-07-08T20:32:06Z

From @khwilliamson

On 07/08/2016 09:57 AM, Graham Knop via RT wrote:

On Fri Jul 08 07:07:08 2016, mail@steffen-mueller.net wrote:

On 07/08/2016 03:58 PM, Peter Rabbitson wrote:

On 07/06/2016 02:39 PM, Peter Rabbitson wrote:

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is
2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time:
https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

Still the case almost 5 days after original upload :/

Sorry, missed that. Been out of order sick for most of that. I'm still
in pretty bad shape now.

Has somebody tried debugging/fixing this?

NB: Normally, I'd say anything as old as 5.12 doesn't require going out
of the way for, but considering virtually everything (usually
inadvertently) depends on DD, this seems worth fixing.

--Steffen

Bisected to this commit:

commit 31ac59b
Author: Karl Williamson <public@khwilliamson.com>
Date: Wed Mar 13 16:16:14 2013 -0600
 Data&#8203;::Dumper&#8203;: Generalize for EBCDIC platforms

 This extends Data&#8203;::Dumper to work on EBCDIC platforms\.  This is just the
 code changes\.  Some \.t files will be changed as well\, in future commits

 This involves some code refactoring especially in the \.xs code to
 collapse EBCDIC/ASCII handling into one\.  The if\-elsif\-elsif\-\.\.\.\-else
 logic is cleaned up\, so that there are fewer branches taken on average\.
---
via perlbug: queue: perl5 status: open
https://rt-archive.perl.org/perl5/Ticket/Display.html?id=128525

I'll investigate

p5pRT · 2016-07-09T00:01:34Z

From @karenetheridge

When releasing, please include this line in Makefile.PL as arguments to WriteMakefile() (as described in https://rt.cpan.org/Ticket/Display.html?id=12282):

+ $] >= 5.005 && $] <= 5.011000 ? ( INSTALLDIRS => 'perl' ) : (),

p5pRT · 2016-07-10T04:07:13Z

From @khwilliamson

On 07/08/2016 09:57 AM, Graham Knop via RT wrote:

On Fri Jul 08 07:07:08 2016, mail@steffen-mueller.net wrote:

On 07/08/2016 03:58 PM, Peter Rabbitson wrote:

On 07/06/2016 02:39 PM, Peter Rabbitson wrote:

On 07/03/2016 12:02 AM, Karen Etheridge (via RT) wrote:

The latest CPAN release of Data::Dumper is 2.154; in blead it is
2.160. blead has been ahead of CPAN since 5.21.6!

Also, this bug must be fixed at the same time:
https://rt.cpan.org/Ticket/Display.html?id=12282

The resolution of the above broke Data::Dumper installs on Perl < 5.14.

Please rectify ASAP.

Still the case almost 5 days after original upload :/

Sorry, missed that. Been out of order sick for most of that. I'm still
in pretty bad shape now.

Has somebody tried debugging/fixing this?

NB: Normally, I'd say anything as old as 5.12 doesn't require going out
of the way for, but considering virtually everything (usually
inadvertently) depends on DD, this seems worth fixing.

--Steffen

Bisected to this commit:

commit 31ac59b
Author: Karl Williamson <public@khwilliamson.com>
Date: Wed Mar 13 16:16:14 2013 -0600
 Data&#8203;::Dumper&#8203;: Generalize for EBCDIC platforms

This bug turns out to be because the isASCII() macro in 5.12 (and I
presume earlier) required its argument to be unsigned (this fact is
undocumented). If you call isASCII in such a release, it always returns
TRUE for a signed argument, as it tests only that the argument < 127,
and so the C optimizer feels free to skip compiling any code in a branch
where it is tested for being false.

I fixed this bug back then, and at the time we got something in ppport.h
to do it correctly. But in looking at that file now, it looks like it's
currently only puts the correct definition if there is no isASCII at all.

It seems to me that the bug should be fixed there; other modules likely
depend on the correct functioning of isASCII, and then no changes need
to be done to D:D except the inclusion of an up-to-date ppport.h.

While we're at it, I have updated definitions for the other isFOO macros
in handy.h, which are constructed so they are valid on EBCDIC as well.
It seems to me these should be used instead of the defintions that are
there, and they probably should override the existing definitions pulled
from the old release's handy.h.

 This extends Data&#8203;::Dumper to work on EBCDIC platforms\.  This is just the
 code changes\.  Some \.t files will be changed as well\, in future commits

 This involves some code refactoring especially in the \.xs code to
 collapse EBCDIC/ASCII handling into one\.  The if\-elsif\-elsif\-\.\.\.\-else
 logic is cleaned up\, so that there are fewer branches taken on average\.

---
via perlbug: queue: perl5 status: open
https://rt-archive.perl.org/perl5/Ticket/Display.html?id=128525

p5pRT · 2016-07-10T13:28:36Z

From @arc

Karl Williamson <public@khwilliamson.com> wrote:

This bug turns out to be because the isASCII() macro in 5.12 (and I presume
earlier) required its argument to be unsigned (this fact is undocumented).

I fixed this bug back then, and at the time we got something in ppport.h to
do it correctly. But in looking at that file now, it looks like it's
currently only puts the correct definition if there is no isASCII at all.

It seems to me that the bug should be fixed there; other modules likely
depend on the correct functioning of isASCII, and then no changes need to be
done to D:D except the inclusion of an up-to-date ppport.h.

Thanks, Karl, that does sound like the right fix.

Until PPPort is updated, Data::Dumper 2.160 can be made to work on
5.12.x (or at least its test suite can be made to pass there) by
patching just one call to isASCII() so that its argument is cast to an
unsigned type. I've attached a patch that does this; it might be
helpful in the short term for people who are using distroprefs or
similar mechanisms.

There's an additional problem on 5.10.0 and earlier (including 5.8.9):
the test suite needs Test::More 0.96, but that dependency is expressed
using BUILD_REQUIRES in Makefile.PL, and therefore isn't honoured by
the EUMM in 5.10.x or 5.8.x. So installing on (say) a vanilla 5.10.0
yields lots of test failures: most of the *.t files try to call
note(), which doesn't exist, so they exit early with an exception.

I've attached a patch against Data-Dumper 2.160 that I think addresses
this, but I'd welcome review from anyone who knows more about
toolchain issues than me. Also, I'm not sure where it should get
applied, since the version of Data::Dumper in core uses an
auto-generated Makefile.PL. Do we just add a new
dist/Data-Dumper/Makefile.PL to blead, with the same contents as the
CPAN version?

--
Aaron Crane ** http://aaroncrane.co.uk/

p5pRT · 2016-07-10T13:28:36Z

From @arc

data_dumper_isascii.patch

diff -ur Data-Dumper-2.160.orig/Dumper.xs Data-Dumper-2.160/Dumper.xs
--- Data-Dumper-2.160.orig/Dumper.xs	2016-07-03 20:08:54.000000000 +0100
+++ Data-Dumper-2.160/Dumper.xs	2016-07-10 11:18:40.000000000 +0100
@@ -369,7 +369,7 @@
             UV k;

             if (do_utf8
-                && ! isASCII(*s)
+                && ! isASCII(*(U8*)s)
                     /* Exclude non-ASCII low ordinal controls.  This should be
                      * optimized out by the compiler on ASCII platforms; if not
                      * could wrap it in a #ifdef EBCDIC, but better to avoid

p5pRT · 2016-07-10T13:28:36Z

From @arc

data_dumper_makefilepl.patch

diff -ur Data-Dumper-2.160.orig/Makefile.PL Data-Dumper-2.160/Makefile.PL
--- Data-Dumper-2.160.orig/Makefile.PL	2013-03-15 09:04:11.000000000 +0000
+++ Data-Dumper-2.160/Makefile.PL	2016-07-10 13:57:06.000000000 +0100
@@ -1,5 +1,9 @@
 use 5.006001;
 use ExtUtils::MakeMaker;
+my $have_test_requires = ExtUtils::MakeMaker->VERSION ge '6.64';
+my %test_req = (
+    'Test::More' => '0.98',
+);
 WriteMakefile(
     NAME             => "Data::Dumper",
     VERSION_FROM     => 'Dumper.pm',
@@ -11,9 +15,10 @@
     MAN3PODS         => {},
     DEFINE           => '-DUSE_PPPORT_H',
     INSTALLDIRS      => 'perl',
-    BUILD_REQUIRES   => {
-        Test::More => '0.98',
+    PREREQ_PM => {
+        $have_test_requires ? () : %test_req,
     },
+    $have_test_requires ? (TEST_REQUIRES => \%test_req) : (),
     META_MERGE => {
         dynamic_config => 0,
         resources => {

p5pRT · 2016-07-11T20:12:39Z

From mail@steffen-mueller.net

Hi Aaron,

On 07/10/2016 03:26 PM, Aaron Crane wrote:

Karl Williamson <public@khwilliamson.com> wrote:

This bug turns out to be because the isASCII() macro in 5.12 (and I presume
earlier) required its argument to be unsigned (this fact is undocumented).

I fixed this bug back then, and at the time we got something in ppport.h to
do it correctly. But in looking at that file now, it looks like it's
currently only puts the correct definition if there is no isASCII at all.

It seems to me that the bug should be fixed there; other modules likely
depend on the correct functioning of isASCII, and then no changes need to be
done to D:D except the inclusion of an up-to-date ppport.h.

Thanks, Karl, that does sound like the right fix.

Until PPPort is updated, Data::Dumper 2.160 can be made to work on
5.12.x (or at least its test suite can be made to pass there) by
patching just one call to isASCII() so that its argument is cast to an
unsigned type. I've attached a patch that does this; it might be
helpful in the short term for people who are using distroprefs or
similar mechanisms.

There's an additional problem on 5.10.0 and earlier (including 5.8.9):
the test suite needs Test::More 0.96, but that dependency is expressed
using BUILD_REQUIRES in Makefile.PL, and therefore isn't honoured by
the EUMM in 5.10.x or 5.8.x. So installing on (say) a vanilla 5.10.0
yields lots of test failures: most of the *.t files try to call
note(), which doesn't exist, so they exit early with an exception.

I've attached a patch against Data-Dumper 2.160 that I think addresses
this, but I'd welcome review from anyone who knows more about
toolchain issues than me. Also, I'm not sure where it should get
applied, since the version of Data::Dumper in core uses an
auto-generated Makefile.PL. Do we just add a new
dist/Data-Dumper/Makefile.PL to blead, with the same contents as the
CPAN version?

Thank you for this!

I've just uploaded DD 2.161 with both patches. I maintain a copy of the
code base in a separate github repo for exactly the Makefile.PL. Other
than your temporary isASCII fix it has no other diffs from blead.

Best regards,
Steffen

p5pRT · 2016-07-12T08:21:58Z

From @ribasushi

On 07/11/2016 10:12 PM, Steffen Mueller wrote:

I've just uploaded DD 2.161 with both patches.

2.161 smoked well on my setup[1] too.

Cheers!

[1] https://travis-ci.org/dbsrgits/dbix-class/builds/142004286

xenu · 2021-05-17T16:14:27Z

A new Data::Dumper release has been uploaded to CPAN a few hours ago.

p5pRT added the Severity Low label Oct 19, 2019

karenetheridge self-assigned this Jan 31, 2020

karenetheridge added the dist-Data-Dumper issues in the dual-life blead-first Data-Dumper distribution label Jan 31, 2020

xenu closed this as completed May 17, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

cpan Data::Dumper is behind blead Data::Dumper #15417

cpan Data::Dumper is behind blead Data::Dumper #15417

p5pRT commented Jul 2, 2016

p5pRT commented Jul 2, 2016

p5pRT commented Jul 6, 2016

p5pRT commented Jul 6, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 8, 2016

p5pRT commented Jul 9, 2016

p5pRT commented Jul 10, 2016

p5pRT commented Jul 10, 2016

p5pRT commented Jul 10, 2016

p5pRT commented Jul 10, 2016

p5pRT commented Jul 11, 2016

p5pRT commented Jul 12, 2016

xenu commented May 17, 2021

cpan Data::Dumper is behind blead Data::Dumper #15417

cpan Data::Dumper is behind blead Data::Dumper #15417

Comments

p5pRT commented Jul 2, 2016

p5pRT commented Jul 2, 2016

From @karenetheridge

p5pRT commented Jul 6, 2016

From @ribasushi

p5pRT commented Jul 6, 2016

p5pRT commented Jul 8, 2016

From @ribasushi

p5pRT commented Jul 8, 2016

From mail@steffen-mueller.net

p5pRT commented Jul 8, 2016

From @ribasushi

p5pRT commented Jul 8, 2016

From mail@steffen-mueller.net

p5pRT commented Jul 8, 2016

From @haarg

p5pRT commented Jul 8, 2016

From @khwilliamson

p5pRT commented Jul 9, 2016

From @karenetheridge

p5pRT commented Jul 10, 2016

From @khwilliamson

p5pRT commented Jul 10, 2016

From @arc

p5pRT commented Jul 10, 2016

From @arc

p5pRT commented Jul 10, 2016

From @arc

p5pRT commented Jul 11, 2016

From mail@steffen-mueller.net

p5pRT commented Jul 12, 2016

From @ribasushi

xenu commented May 17, 2021