From @steve-m-hay

When building blead with MinGW-w64-4.8.0 I see hundreds of lines like this come out from the Errno build​:

Bareword found where operator expected at (eval 22) line 1, near "13834l"
  (Missing operator before l?)

This doesn't happen with MinGW-w64-4.7.4 or earlier, or MinGW, or VC++.

The change in behaviour is due to changes in the MinGW-w64 headers. E.g. 4.7.4 has​:

#define ERROR_IPSEC_IKE_PROCESS_ERR_ID 13834L

but now 4.8.0 has​:

#define ERROR_IPSEC_IKE_PROCESS_ERR_ID __MSABI_LONG(13834)
#define __MSABI_LONG(x) x ## l

The lowercase "l" rather than uppercase "L" trips up Errno_pm.PL. The attached patch fixes this, and I think it should be a 5.20 blocker -- as the attached Errnos.tar.gz shows, the difference between the 4.7.4-generated Errno.pm and the original (i.e. current) 4.8.0-generated Errno.pm is HUGE. The fixed 4.8.0-generated Errno.pm is much more like it.

From @steve-m-hay

diff --git a/ext/Errno/Errno_pm.PL b/ext/Errno/Errno_pm.PL
index db8ada7..55ad01a 100644
--- a/ext/Errno/Errno_pm.PL
+++ b/ext/Errno/Errno_pm.PL
@@ -2,7 +2,7 @@ use ExtUtils::MakeMaker;
 use Config;
 use strict;
 
-our $VERSION = "1.20_02";
+our $VERSION = "1.20_03";
 
 my %err = ();
 
@@ -250,7 +250,7 @@ sub write_errno_pm {
 	    next unless ($name, $expr) = /"(.*?)"\s*\[\s*\[\s*(.*?)\s*\]\s*\]/;
 	    next if $name eq $expr;
 	    $expr =~ s/\(?\(\s*[a-z_]\w*\s*\)([^\)]*)\)?/$1/i; # ((type)0xcafebabe) at alia
-	    $expr =~ s/((?:0x)?[0-9a-fA-F]+)[LU]+\b/$1/g; # 2147483647L et alia
+	    $expr =~ s/((?:0x)?[0-9a-fA-F]+)[luLU]+\b/$1/g; # 2147483647L et alia
 	    next if $expr =~ m/^[a-zA-Z]+$/; # skip some Win32 functions
 	    if($expr =~ m/^0[xX]/) {
 		$err{$name} = hex $expr;

From @steve-m-hay

Errnos.tar.gz

From @steve-m-hay

Applied in commit f974e9b.

@steve-m-hay - Status changed from 'new' to 'resolved'