Bleadperl v5.22.0-131-gaa8f6ce breaks GFUJI/App-test-travis-v0.9.7.tar.gz #15035
Comments
From @andkbisect commit aa8f6ce Microoptimize some matches in utf8_heavy.pl cpantesters http://www.cpantesters.org/cpan/report/a496ba7c-357b-11e5-be3b-567b92740143 discovered by slaven: gfx/App-test-travis#6 perl -V Summary of my perl5 (revision 5 version 23 subversion 1) configuration: Characteristics of this binary (from libperl): -- |
From @tonycozOn Sun Nov 08 22:27:21 2015, andreas.koenig.7os6VVqR@franz.ak.mind.de wrote:
App::test::travis loads encoding::warnings which sets ${^ENCODING}. When ${^ENCODING} is set S_scan_str() (toke.c) treats the match and replacement strings as unicode for this line: $VERSION =~ tr/_//d; in Carp.pm, S_pmtrans() (op.c) sees that they're SvUTF8() and rather than using the simple lookup table you'd expect for such a simple tr/// it builds a swash definition and attempts to create a new swash by calling utf8::SWASHNEW. This in turn calls the utf8::AUTOLOAD, which loads utf8_heavy.pl, which compiles until this line: (my $loose = $_[0]) =~ tr/-_ \t//d; At this point both Carp.pm and utf8_heavy.pl are only partly loaded, and again S_pmtrans sees UTF8 marked match and replacement strings and again attempts to all utf8::SWASHNEW, again falling back to utf8::AUTOLOAD. This time around utf_heavy.pl and Carp.pm are already marked as loading, so the requires: sub AUTOLOAD { are skipped and we see an error for the missing croak() definition. Possible fixes: 1) modify S_scan_str() to attempt to downgrade the string when IN_ENCODING This is probably the correct fix, if ${^ENCODING} wasn't deprecated. 2) add C<< local ${^ENCODING}; >> to the top of utf8::AUTOLOAD A simple workaround. 3) Finally remove the deprecated ${^ENCODING}. 4) Partly revert aa8f6ce (leaving a mine for someone else to step on in the future) Tony |
|
The RT System itself - Status changed from 'new' to 'open' |
From @khwilliamsonaa8f6ce |
From @khwilliamsonOn 11/11/2015 10:22 PM, Tony Cook via RT wrote:
I haven't looked at this in detail, but this seems to me to likely be I think that the problem is that tr/// can load utf8_heavy.pl which We could write perhaps a save_tr_context(), but save_re_context() is It seems to me to be the best short term thing to forbid tr/// in Longer term, it would be better to not have tr/// use swashes at all, Even longer term is getting rid of utf8_heavy altogether. Aaron Crane I looked last year at the amount of work involved in getting rid of 'use
|
From @rgarciaOn 12 November 2015 at 19:16, Karl Williamson <public@khwilliamson.com> wrote:
I have hard evidence of the time difference, but it's relatively tiny. I'm
I suspect lots of cargo culting in usage of encoding.pm. |
From @tonycozOn Thu Nov 12 10:25:43 2015, public@khwilliamson.com wrote:
Perhaps something like the attached. Tony |
From @tonycoz0001-perl-126593-make-sure-utf8_heavy.pl-doesn-t-depend-o.patchFrom ca0f8d3b24deeaf0fb8e0297ec69a3614a0e2143 Mon Sep 17 00:00:00 2001
From: Tony Cook <tony@develop-help.com>
Date: Thu, 26 Nov 2015 16:22:04 +1100
Subject: [perl #126593] make sure utf8_heavy.pl doesn't depend on itself
With ${^ENCODING} set, it did.
Partly reverts:
commit aa8f6cef961dc2009604f7464c66106421c3ae81
Author: Rafael Garcia-Suarez <rgs@consttype.org>
Date: Wed Jun 17 13:18:59 2015 +0200
Microoptimize some matches in utf8_heavy.pl
---
MANIFEST | 1 +
lib/utf8_heavy.pl | 2 +-
t/uni/heavy.t | 40 ++++++++++++++++++++++++++++++++++++++++
3 files changed, 42 insertions(+), 1 deletion(-)
create mode 100644 t/uni/heavy.t
diff --git a/MANIFEST b/MANIFEST
index ca2c455..b10f84f 100644
--- a/MANIFEST
+++ b/MANIFEST
@@ -5550,6 +5550,7 @@ t/uni/fold.t See if Unicode folding works
t/uni/goto.t See if Unicode goto &sub works
t/uni/greek.t See if Unicode in greek works
t/uni/gv.t See if Unicode GVs work.
+t/uni/heavy.t See if utf8_heavy.pl uses perl that depends on it
t/uni/labels.t See if Unicode labels work
t/uni/latin2.t See if Unicode in latin2 works
t/uni/lex_utf8.t See if Unicode in lexer works
diff --git a/lib/utf8_heavy.pl b/lib/utf8_heavy.pl
index 0d2e662..872704a 100644
--- a/lib/utf8_heavy.pl
+++ b/lib/utf8_heavy.pl
@@ -20,7 +20,7 @@ sub _loose_name ($) {
# out blanks, underscores and dashes. The complication stems from the
# grandfathered-in 'L_', which retains a single trailing underscore.
- (my $loose = $_[0]) =~ tr/-_ \t//d;
+ (my $loose = $_[0]) =~ s/[-_ \t]//g;
return $loose if $loose !~ / ^ (?: is | to )? l $/x;
return 'l_' if $_[0] =~ / l .* _ /x; # If original had a trailing '_'
diff --git a/t/uni/heavy.t b/t/uni/heavy.t
new file mode 100644
index 0000000..c257dbc
--- /dev/null
+++ b/t/uni/heavy.t
@@ -0,0 +1,40 @@
+#!./perl -w
+# tests that utf8_heavy.pl doesn't use anything that prevents it loading
+BEGIN {
+ chdir 't' if -d 't';
+ @INC = '../lib';
+ require './test.pl';
+}
+
+plan tests => 1;
+
+# see [perl #126593]
+fresh_perl_is(<<'EOP', "", { stderr => 1 }, "doesn't break with \${^ENCODING}");
+no warnings qw(deprecated);
+package Foo;
+sub cat_decode {
+ # stolen from Encode.pm
+ my ( undef, undef, undef, $pos, $trm ) = @_;
+ my ( $rdst, $rsrc, $rpos ) = \@_[ 1, 2, 3 ];
+ use bytes;
+ if ( ( my $npos = index( $$rsrc, $trm, $pos ) ) >= 0 ) {
+ $$rdst .=
+ substr( $$rsrc, $pos, $npos - $pos + length($trm) );
+ $$rpos = $npos + length($trm);
+ return 1;
+ }
+ $$rdst .= substr( $$rsrc, $pos );
+ $$rpos = length($$rsrc);
+ return q();
+}
+
+sub decode {
+ my (undef, $tmp) = @_;
+ utf8::decode($tmp);
+ $tmp;
+}
+
+BEGIN { ${^ENCODING} = bless [], q(Foo) };
+
+(my $tmp = q(abc)) =~ tr/abc/123/;
+EOP
--
2.1.4
|
From @tonycozOn Wed Nov 25 21:25:25 2015, tonyc wrote:
Pushed as 0fd86aa. Tony |
|
@tonycoz - Status changed from 'open' to 'pending release' |
From @khwilliamsonThank you for submitting this report. You have helped make Perl better. Perl 5.24.0 may be downloaded via https://metacpan.org/release/RJBS/perl-5.24.0 |
|
@khwilliamson - Status changed from 'pending release' to 'resolved' |
Migrated from rt.perl.org#126593 (status was 'resolved')
Searchable as RT126593$
The text was updated successfully, but these errors were encountered: