From perl@cjmweb.net

I tried to build Rakudo Star 2010.07, but it failed with

  src/gen/core.pm > src/gen/core.pir
===SORRY!===
Symbol '@​named-params\ufffd' not predeclared in MAIN_HELPER
(src/gen/core.pm​:6542)
make​: *** [perl6.pbc] Error 1

I'm running 32-bit Gentoo Linux with Perl 5.10.1.

I investigated and found that the offending line in src/gen/core.pm
contained

  @​named-params».name».substr(1)

But the original line in src/core/MAIN.pm contained

  @​named-params».name».substr(1)

"»" in UTF-8 is C2 BB, which in ISO-8859-1 is "»".

So it appears that MAIN.pm is being read as if it were ISO-8859-1 and
then written out in UTF-8. I'm not sure why that happens, but I
discovered that removing the generated files, "unset PERL_UNICODE", and
running "make" again allowed the build to complete.

I had PERL_UNICODE set to SAL, and that causes the build to fail
consistently. I don't know why.

Steps to reproduce​:

Unpack rakudo-star-2010.07.tar.gz
  $ perl Configure.pl --gen-parrot
  $ PERL_UNICODE=SAL make

--
Chris Madsen perl@​cjmweb.net
  -------------------- http://www.cjmweb.net --------------------

From perl@cjmweb.net

I forgot to mention that my LANG=en_US.utf8, and no LC_* vars are set.

perl@cjmweb.net - Status changed from 'new' to 'open'

perl@cjmweb.net - Status changed from 'open' to 'new'

From perl@cjmweb.net

build/gen_core_pm.pl is opening files using

  open(my $fh, "<", $file) or die $!;

but writing to STDOUT. My PERL_UNICODE setting causes STDOUT to be
encoded as UTF-8, but the default encoding for open remains ISO-8859-1.

gen_core_pm.pl (and similar scripts) need to either binmode STDOUT, or
open files with the correct encoding.

From perl@cjmweb.net

The attached patch (from
http://github.com/madsen/rakudo/commit/450fccceb8a3963202f6f5783f83b56449dcc219
) appears to fix this problem. gen_core_pm.pl was the only file I
noticed that didn't specify the encoding for I/O.

From perl@cjmweb.net

From 450fccceb8a3963202f6f5783f83b56449dcc219 Mon Sep 17 00:00:00 2001
From: Christopher J. Madsen <perl@cjmweb.net>
Date: Sat, 31 Jul 2010 10:49:19 -0500
Subject: [PATCH] Explicitly request :utf8 layer in gen_core_pm.pl to fix RT #76856

---
 build/gen_core_pm.pl |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/build/gen_core_pm.pl b/build/gen_core_pm.pl
index 10208ad..ce4198e 100644
--- a/build/gen_core_pm.pl
+++ b/build/gen_core_pm.pl
@@ -4,6 +4,9 @@
 
 use strict;
 use warnings;
+use 5.008;
+
+binmode STDOUT, ':utf8';
 
 my @files = @ARGV;
 
@@ -15,7 +18,7 @@ END_SETTING
 my %classnames;
 foreach my $file (@files) {
     print "# From $file\n\n";
-    open(my $fh, "<",  $file) or die $!;
+    open(my $fh, "<:utf8",  $file) or die "$file: $!";
     local $/;
     my $x = <$fh>;
     close $fh;
-- 
1.6.4.4

From @moritz

On Sat Jul 31 08​:56​:19 2010, cjm wrote​:

The attached patch (from

http://github.com/madsen/rakudo/commit/450fccceb8a3963202f6f5783f83b56449dcc219

) appears to fix this problem. gen_core_pm.pl was the only file I
noticed that didn't specify the encoding for I/O.

Thank you very much, applied.

Cheers,
Moritz

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

@moritz - Status changed from 'open' to 'resolved'