Skip Menu |

This queue is for tickets about the HTML-Mason CPAN distribution.

Report information
The Basics
Id: 121443
Status: resolved
Priority: 0/
Queue: HTML-Mason

People
Owner: Nobody in particular
Requestors: KENTNL [...] cpan.org
Cc:
AdminCc:

Bug Information
Severity: (no value)
Broken in: (no value)
Fixed in: (no value)



Subject: t/13-errors.t fails when no '.' in @INC (not ok 26 - require_bad_module_in_once )
Note: templated response follows

Perl 5.25.11 removes '.' from the default @INC, and this breaks `require "t/lib/BadModule.pm"` due to not being able to assume ./t/lib/BadModule.pm is in the @INC load path.

This is not a complete end-of-the-world problem, as Perl provides a temporary workaround variable (PERL_USE_UNSAFE_INC=1) affected users can set that restores '.' in @INC temporarily to "just work", and for end users convenience, CPAN.pm, cpanm, prove and TAP::Harness ( called in `make test` ) set this variable, creating an illusion during installation that this problem is not there

However, this should still be considered a serious issue, as many types of end users don't get the luxury afforded by the CPAN-and-friends hacks ( vendors, hand installers, people running `perl t/foo.t` ), and this workaround is scheduled to be removed in the future.

And due to aforementioned temporary conveniences and hacks, you should intentionally defeat said hacks when testing you've fixed this, by explicitly setting PERL_USE_UNSAFE_INC=0 in your environment prior to running Makefile.PL and tests, and ensuring to perform said tests on Perl 5.25.11 or newer

Two primary solutions are recommended for tests:

1. Re-organise test dependencies into a dedicated directory such as `t/lib` and load that path into @INC with `use lib 't/lib'`

2. Rewrite calls to `do` and `require` to have either absolute paths ( via `File::Spec rel2abs` ) or have explicit `./` prefixes, both of which disable @INC lookup

Both of these suggestions intentionally avoid the easier option of just reinserting `.` on purpose, as some proportion of runtime code also is subject to these cwd-implied bugs, and it is desirable to expose such cases as the errors they are, not silently suppress their failures

For more details, check http://blogs.perl.org/users/ryan_voots/2017/04/trials-and-troubles-with-changing-inc.html

---

Raw Error Observed:

perl -Ilib t/13-errors.t
1..26
ok 1 - bad_args
ok 2 - backtrace
ok 3 - read_zero_size
ok 4 - bad_source_callback
ok 5 - bad_escape_flag
ok 6 - error_mode_output
ok 7 - error_in_subrequest
ok 8 - check_error_format
ok 9 - check_exec_not_found
ok 10 - check_exec_not_found_html_format
ok 11 - check_comp_not_found
ok 12 - change_error_format
ok 13 - check_error_format_brief
ok 14 - object_exception
ok 15 - subcomponent_redefined
ok 16 - method_redefined
ok 17 - method_subcomp_conflict
ok 18 - subcomp_bad_name
ok 19 - content_comp_wrong_error
ok 20 - top_level_compilation_error
ok 21 - component_error_handler_false
ok 22 - component_error_Handler_no_upgrade
ok 23 - component_error_handler_false_fatal_mode
ok 24 - component_error_handler_uc_message
ok 25 - use_bad_module
not ok 26 - require_bad_module_in_once
#   Failed test 'require_bad_module_in_once'
#   at lib/HTML/Mason/Tests.pm line 553.
# Looks like you failed 1 test of 26.

perl -Mlib=. -Ilib t/13-errors.t
1..26
ok 1 - bad_args
ok 2 - backtrace
ok 3 - read_zero_size
ok 4 - bad_source_callback
ok 5 - bad_escape_flag
ok 6 - error_mode_output
ok 7 - error_in_subrequest
ok 8 - check_error_format
ok 9 - check_exec_not_found
ok 10 - check_exec_not_found_html_format
ok 11 - check_comp_not_found
ok 12 - change_error_format
ok 13 - check_error_format_brief
ok 14 - object_exception
ok 15 - subcomponent_redefined
ok 16 - method_redefined
ok 17 - method_subcomp_conflict
ok 18 - subcomp_bad_name
ok 19 - content_comp_wrong_error
ok 20 - top_level_compilation_error
ok 21 - component_error_handler_false
ok 22 - component_error_Handler_no_upgrade
ok 23 - component_error_handler_false_fatal_mode
ok 24 - component_error_handler_uc_message
ok 25 - use_bad_module
ok 26 - require_bad_module_in_once
 

-- 
- CPAN kentnl@cpan.org
- Gentoo Perl Maintainer kentnl@gentoo.org ( perl@gentoo.org )
Subject: 0001-Fix-t-13-errors.t-failing-w-o-.-in-INC.patch
From 56ff714ab4daeaeed29664ddb7b43db13dcf5e57 Mon Sep 17 00:00:00 2001 From: Kent Fredric <kentnl@gentoo.org> Date: Tue, 24 Oct 2017 17:40:13 +1300 Subject: Fix t/13-errors.t failing w/o '.' in @INC Bug: https://bugs.gentoo.org/616966 Bug: https://rt.cpan.org/Public/Bug/Display.html?id=121443 --- t/13-errors.t | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/t/13-errors.t b/t/13-errors.t index 716f2f02..78cfc6b0 100644 --- a/t/13-errors.t +++ b/t/13-errors.t @@ -417,7 +417,7 @@ EOF description => 'Require a module with an error in a once block', component => <<'EOF', <%once> -require "t/lib/BadModule.pm"; +require "./t/lib/BadModule.pm"; </%once> hi! EOF -- 2.14.2