Fix to build on Perl 5.37.9 #5

demerphq · 2023-02-28T11:11:22Z

Fixes the test failure, silences "not imported" warnings during test, and bumps version to v0.14.

This also includes changes to Makefile.PL so that it depends on EUMM 7.66 (the latest), and removes the META.yml which was resulting in bogus warnings about an unknown license and bizarre MYMETA.yml contents even when run against a modern EUMM. It also simplifies the Makefile.PL to just depend on a modern EUMM instead of doing different stuff for older ones. AFAIK, EUMM is backwards compatible to perl 5.8.0 so there is no reason not to expect it to be up to date.

The test failure is related to: Perl/perl5#20357

Fixes: Perl/perl5#20864

…rors In Perl/perl5#20357 the Perl dev team fixed some inconsistencies in how eval behaved with regard to compile errors. Prior that PR the exact outcome of a compile error varied with the type and number of errors that were encountered, sometimes calling $SIG{__DIE__}, sometimes not, sometimes outputing "Execution ... aborted" messages, sometimes not. With the merge of that PR the behavior is now consistent, and in particular for this module the "Execution ... aborted" message is now always output when there is a compile error. This patch strips this message off the error before rethrowing the exception with _croak(). This patch also includes some minor changes to the code to make it easier to debug the code being eval()ed which I left commented out. I figured if it was useful to debug this issue then it might useful again in the future.

…stinct var names Using "$FILE" in multiple tests causes the "not imported" warnings. Simply replacing them with distinct names causes the warnings to go away.

Leont · 2023-03-01T01:02:12Z

Makefile.PL

@@ -28,15 +16,11 @@ WriteMakefile(
 	'Slay::Maker' => 0.04,
 	'File::Path'  => 0,
 	'File::Copy::Recursive' => 0,
+        'ExtUtils::MakeMaker' => 7.66,


Why require such a recent MakeMaker?

Because the makfile does a bunch of contortions to deal with older makefiles, a simpler solution is to just require a modern one.

That doesn't explain why it would need more than the 6.54 that was previously suggested

Leont · 2023-03-01T01:04:41Z

I'm puzzled by it removing the META.yml file; I can't imagine that being the solution to any problem.

demerphq · 2023-03-01T06:42:17Z

Removing it stopped the complaint about the license.

demerphq · 2023-03-01T07:50:44Z

I'm puzzled by it removing the META.yml file; I can't imagine that being the solution to any problem.

Shrug. Your imagination is faulty I guess. [ What an arrogant comment. Sheesh. ]

All of the below is assuming that I have removed the commit which alters the Makefile.PL and the META.yml in this PR, that is: 5dd2c2f

With the patch for EUMM from Fix strange warning from existing META.yml license info conflicting with Makefile.PL Perl-Toolchain-Gang/ExtUtils-MakeMaker#436 and with the META.yaml I get this:

$ perl Makefile.PL 
Generating a Unix-style Makefile
Writing Makefile for Slay::Makefile
Your META.yml has a different license (unknown) than your Makefile.PL (perl_5)
Writing MYMETA.yml and MYMETA.json

With the latest version of EUMM in the repo, and with the META.yml:

$ perl Makefile.PL 
Generating a Unix-style Makefile
Writing Makefile for Slay::Makefile
Invalid LICENSE value 'perl_5' ignored
Writing MYMETA.yml and MYMETA.json

Removing the META.yml:

$ rm META.yml
$ perl Makefile.PL 
Generating a Unix-style Makefile
Writing Makefile for Slay::Makefile
Writing MYMETA.yml and MYMETA.json

Difference in the MYMETA.yml and the META.yml from the last step:

$ mv MYMETA.yml META.yml
$ git diff
diff --git a/META.yml b/META.yml
index a7834b5..f7e7b58 100644
--- a/META.yml
+++ b/META.yml
@@ -1,16 +1,29 @@
---- #YAML:1.0
-name:                Slay-Makefile
-version:             0.12
-abstract:            Wrapper to Slay::Maker that reads the rules from a file
-license:             ~
-author:              
-    - Mark Nodine <[email protected]>
-generated_by:        ExtUtils::MakeMaker version 6.44
-distribution_type:   module
-requires:     
-    File::Copy::Recursive:         0
-    File::Path:                    0
-    Slay::Maker:                   0.04
+---
+abstract: 'Wrapper to Slay::Maker that reads the rules from a file'
+author:
+  - 'Slaven Rezic <[email protected]>'
+  - 'Mark Nodine <[email protected]>'
+  - 'Mark Nodine <[email protected]>'
+build_requires:
+  ExtUtils::MakeMaker: '0'
+configure_requires:
+  ExtUtils::MakeMaker: '0'
+dynamic_config: 0
+generated_by: 'ExtUtils::MakeMaker version 7.66, CPAN::Meta::Converter version 2.150010'
+license: perl
 meta-spec:
-    url:     http://module-build.sourceforge.net/META-spec-v1.3.html
-    version: 1.3
+  url: http://module-build.sourceforge.net/META-spec-v1.4.html
+  version: '1.4'
+name: Slay-Makefile
+no_index:
+  directory:
+    - t
+    - inc
+requires:
+  File::Copy::Recursive: '0'
+  File::Path: '0'
+  Slay::Maker: '0.04'
+resources:
+  repository: git://github.com/eserte/Slay-Makefile.git
+version: '0.13'
+x_serialization_backend: 'CPAN::Meta::YAML version 0.018'

Observable differences seem to be,
a) the license is more correct. I do not know why perl_5 turned into perl, but surely that is more correct than ~.
b) Mark Nodine <[email protected]> is mentioned twice instead of once. Presumably because of this:

($eumm_recent_enough
     ? (META_MERGE => { resources  => { repository => 'git://github.com/eserte/Slay-Makefile.git' },
                        author     => ['Mark Nodine <[email protected]>'],
                      },
       )
     : ()
    ),

Checking out the code at yves/fix_for_perl_5_37_9 (that is restoring 5dd2c2f) and applying a fixup on top (i shouldnt have removed the author line from the META_MERGE):

perl Makefile.PL
Generating a Unix-style Makefile
Writing Makefile for Slay::Makefile
Writing MYMETA.yml and MYMETA.json

Notice there is complaint about an invalid license.

And reviewing the final MYMETA.yml:

$ cat MYMETA.yml 
---
abstract: 'Wrapper to Slay::Maker that reads the rules from a file'
author:
  - 'Slaven Rezic <[email protected]>'
  - 'Mark Nodine <[email protected]>'
  - 'Mark Nodine <[email protected]>'
build_requires:
  ExtUtils::MakeMaker: '0'
configure_requires:
  ExtUtils::MakeMaker: '0'
dynamic_config: 0
generated_by: 'ExtUtils::MakeMaker version 7.66, CPAN::Meta::Converter version 2.150010'
license: perl
meta-spec:
  url: http://module-build.sourceforge.net/META-spec-v1.4.html
  version: '1.4'
name: Slay-Makefile
no_index:
  directory:
    - t
    - inc
requires:
  ExtUtils::MakeMaker: '7.66'
  File::Copy::Recursive: '0'
  File::Path: '0'
  Slay::Maker: '0.04'
resources:
  repository: git://github.com/eserte/Slay-Makefile.git
version: '0.14'
x_serialization_backend: 'CPAN::Meta::YAML version 0.018'

Conclusion, META_MERGE is buggy (why is Mark added twice?), and keeping the current META.yml isn't helpful particularly. IMO @eserte would be better off removing it, regenerating a MYMETA.yml and then copying that in place.

demerphq · 2023-04-16T10:54:42Z

Note this was reported again as Perl/perl5#21026 which I have closed as we still have Perl/perl5#20864

The failed AppVeyor tests seem to be due to a misconfiguration and are irrelevant to this PR.

Leont · 2023-04-16T12:09:24Z

Removing it stopped the complaint about the license.

But when you «make dist», it will recreate the META.yml in the actual tarball, so any issue would reappear.

The META.yml just does the wrong thing so remove it. Remove the logic to handle older ExtUtls::MakeMaker and instead just require a modern EUMM to proceed.

demerphq · 2023-04-16T12:52:44Z

@Leont My primary intention of this PR was to fix the issue caused by my changes to Perl 5.37.9 so we could clear the related BBC ticket. The change to the Makefile.PL is a secondary feature because I wanted to stop it from reporting "Invalid LICENSE value 'perl_5' ignored", but since you keep complaining about it I have created #6 without that commit and I will close this PR. Feel free to post your own patch to fix the "invalid license" issue if you want.

demerphq · 2023-04-16T13:09:25Z

Removing it stopped the complaint about the license.

But when you «make dist», it will recreate the META.yml in the actual tarball, so any issue would reappear.

Well, I just tried it, and no, the issue does not reappear in the tarball. I have not dug into why, but I think it is because in the current META.yaml contains:

-license:             ~, but the new one contains

but the new one (recreated as you stated above) contains:

license: perl

demerphq added 2 commits February 28, 2023 10:40

t/10_slaymakefile.init/ - Silence "not imported" warnings by using di…

eff95bd

…stinct var names Using "$FILE" in multiple tests causes the "not imported" warnings. Simply replacing them with distinct names causes the warnings to go away.

demerphq mentioned this pull request Feb 28, 2023

BBC: Bleadperl breaks eval error message tests by adding "Execution of ... aborted due to compilation errors." to compile errors. Perl/perl5#20864

Closed

demerphq force-pushed the yves/fix_for_perl_5_37_9 branch from a0c1328 to 5dd2c2f Compare February 28, 2023 13:14

Leont reviewed Mar 1, 2023

View reviewed changes

demerphq mentioned this pull request Apr 16, 2023

BBC: Blead breaks SREZIC/Slay-Makefile-0.13.tar.gz Perl/perl5#21026

Closed

Slay/Makefile.pm - Bump version to 0.14, remove META.yml, update deps.

61bca8e

The META.yml just does the wrong thing so remove it. Remove the logic to handle older ExtUtls::MakeMaker and instead just require a modern EUMM to proceed.

demerphq force-pushed the yves/fix_for_perl_5_37_9 branch from 851d749 to 61bca8e Compare April 16, 2023 12:34

demerphq mentioned this pull request Apr 16, 2023

Fixes for Perl 5.37.9 but no version bump. (Deal with eval exception text change) #6

Open

demerphq closed this Apr 16, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix to build on Perl 5.37.9 #5

Fix to build on Perl 5.37.9 #5

demerphq commented Feb 28, 2023 •

edited

Loading

Leont Mar 1, 2023

demerphq Mar 1, 2023

Leont Apr 16, 2023

Leont commented Mar 1, 2023

demerphq commented Mar 1, 2023

demerphq commented Mar 1, 2023

demerphq commented Apr 16, 2023

Leont commented Apr 16, 2023

demerphq commented Apr 16, 2023

demerphq commented Apr 16, 2023

Fix to build on Perl 5.37.9 #5

Fix to build on Perl 5.37.9 #5

Conversation

demerphq commented Feb 28, 2023 • edited Loading

Leont Mar 1, 2023

Choose a reason for hiding this comment

demerphq Mar 1, 2023

Choose a reason for hiding this comment

Leont Apr 16, 2023

Choose a reason for hiding this comment

Leont commented Mar 1, 2023

demerphq commented Mar 1, 2023

demerphq commented Mar 1, 2023

demerphq commented Apr 16, 2023

Leont commented Apr 16, 2023

demerphq commented Apr 16, 2023

demerphq commented Apr 16, 2023

demerphq commented Feb 28, 2023 •

edited

Loading