Wise Monks,

I am doing an in-place replacement in a text file which succeeds but one (?) extra newline is added at the end of the file:

The input file (a):

1
2
3
[download]

The command:

Output

1
4
3

# there is an extra newline at the end, above
[download]

Looping the above command (the replacement is done only the first time), will yield as many newlines at the end:

for i in $(seq 1 10); do perl -i.x -lp0e 's/2/4/gs' a; done
[download]

bw, bliako

Dear Monks,

I can not understand why this works:

my $x = <<'EOX';
A {
  B 123
  C xyz
}
EOX
$x =~ s/A\s*\{.+?\}//s;
print "x='$x'\n"; # empty
[download]

but this does not (linux terminal):

perl -pe 's/A\s*\{.+?\}//s' <<'EOX'
A {
  B 123
  C xyz
}
EOX
# prints exactly what the input is
[download]

I have checked if there are mistaken shell interpolations to the oneliner script, but there are not, I think:

x='s/A\s*\{.+?\}//s'
echo $x
# s/A\s*\{.+?\}//s
[download]

Edit: the same problem if the oneliner operates on a file:

cat > infile<<'EOX'
A {
  B 123
  C xyz
}
EOX
perl -pe 's/A\s*\{.+?\}//s' infile
# or perl -i -pe 's/A\s*\{.+?\}//s' infile
[download]

Is there any IPv6-capable "paranoid" DNS resolver like Net::DNS::Paranoid, except with support for IPv6?

I'm currently writing a "link preview" fetcher, using Mojo::UserAgent. As this preview fetcher mildly implies fetching user-generated and/or user-controlled links (even though the application is primarily for myself), I'm looking for something like LWPx::ParanoidAgent / LWP::UserAgent::Paranoid / HTTP::Tiny::Paranoid, but for Mojolicious.

I have already resolved (hah!) myself to writing Mojo::UserAgent::Paranoid, which delegates (as they all do) the DNS resolution to Net::DNS::Paranoid, a DNS resolver class that blocks hosts and IP addresses that are unlikely URLs for external services, especially localhost and other internal networks (192.168.*, 10.* and the multicast addresses).

This is all fine and good, if it weren't for IPv6. Net::DNS::Paranoid happily lets ::1 through, as the code mostly looks at IPv4 addresses and its DNS resolution also only speaks A records, not AAAA.

Simply adding ::1 (localhost in IPv6-speak) to the list of blocked hosts is a good start, but I would also want to block the multicast and link-local IPv6 addresses and some others. So, I'm mildly looking towards extending Net::DNS::Paranoid to also support IPv6, but maybe somebody has already done the work, or has links to an existing implementation in another language where I can crib the pitfalls and footguns from.

OS=Linux

I'm struggling to send special characters like umlauts ("äöüÄÖÜ") via X11::GUITest and SendKeys() to another application.

Is there a trick I am missing? Anything after the first special character is discarded.

If I try the linux tool xte from the package 'xautomation', it works fine from the console:

gnome-text-editor & sleep 1; xte "str äöüÄÖÜß"

2 C functions below differ in a single line marked with 3 stars. I thought that a modern decent compiler won't notice. I'm afraid I'm not qualified to analyse compiler flags differences. It's Win10, and 5.32 set-up long ago and portable 5.42 (PDL-edition) unpacked just now. (The code is about PWC-342-2, it's irrelevant and doesn't matter if it still returns valid answers after reduction to SSCCE). Have compiler optimisation settings been changed, in recent Strawberries, in what looks like wrong direction? (+ some more strange issues, but I'd better stay focused on this one)

use strict;
use warnings;
use feature 'say';
use Benchmark 'cmpthese';

use constant L => 1e5;
my $str = '1' x L;
substr $str, rand L, 1 , '0' for 1 .. L;

use Inline C => << 'END_OF_C';

int test_c_3( SV* str ) {
    int i, tmp, acc, max;
    
    STRLEN len;
    char* buf = SvPVbyte( str, len );
    len --;
    
    acc = 0;
    max = -2;
    for( i = 0; i < len; i ++ ) {
        acc += ( buf[ i ] == '1' );         // ***
        tmp = i - 2 * acc;
        if ( tmp > max ) max = tmp;
    }
    
    return max + acc + ( buf[ len ] == '1' ) + 1;
}

int test_c_4( SV* str ) {
    int i, tmp, acc, max;
    
    STRLEN len;
    char* buf = SvPVbyte( str, len );
    len --;
    
    acc = 0;
    max = -2;
    for( i = 0; i < len; i ++ ) {
        if ( buf[ i ] == '1' ) acc ++;      // ***
        tmp = i - 2 * acc;
        if ( tmp > max ) max = tmp;
    }
    
    return max + acc + ( buf[ len ] == '1' ) + 1;
}

END_OF_C

say $^V;
say 'String length: ', L;

cmpthese -2, {
    c3 => sub { test_c_3( $str )},
    c4 => sub { test_c_4( $str )},
};

__END__

v5.32.1
String length: 100000
     Rate  c4  c3
c4 8800/s  -- -0%
c3 8817/s  0%  --

v5.42.0
String length: 100000
     Rate   c4   c3
c4 2511/s   -- -71%
c3 8755/s 249%   --
[download]

Here we go again, a pretty little problem with installing a CPAN module that I want to be on my system. The failure is with Module::Manifest::Skip, its tests won't get past the compile stage. The relevant message as appears below is Can't locate object method "_install_subs" via package "Module::Manifest::Skip". The failure was thrown at /usr/share/perl5/vendor_perl/5.40/Moo.pm line 49, so I updated Moo to release 2.005005 but that didn't solve the problem.

My system has these characteristics:

/usr/bin/perl    v5.40.3
MS Windows 11
CYGWIN_NT-10.0-26100 3.6.4-1.x86_64 2025-07-15 07:55 UTC x86_64 Cygwin
@INC:
/usr/local/lib/perl5/site_perl/5.40/x86_64-cygwin-threads
/usr/local/share/perl5/site_perl/5.40
/usr/lib/perl5/vendor_perl/5.40/x86_64-cygwin-threads
/usr/share/perl5/vendor_perl/5.40
/usr/lib/perl5/5.40/x86_64-cygwin-threads
/usr/share/perl5/5.40

The following console output is a bit redundant but I wanted to be thorough:

Oct 12, 2025 at 18:37 UTC

use strict;
use Excel::Writer::XLSX;
 use Spreadsheet::ParseXLSX;
use Carp qw( croak );
use Cwd qw( abs_path );
use Path::Class;
 use Win32::OLE qw(in);
use Win32::OLE::Const 'Microsoft Word';
...

my $path=....;

    my $word =Win32::OLE->new('Word.Application','Quit');   
     my $word_file = file(abs_path($path)); 
 
    my $doc = $word->{Documents}->Open("$word_file");
print "after doc\n";
    my $tables = $word->ActiveDocument->{Tables};
 
    for my $table (in $tables) {
 
my $numrows=$table->Rows->Count; 
 $main::numcols=$table->Columns->Count; 
 
# Iterate through rows and count columns
foreach my $rownum (1 .. $table->Rows->Count) {
    my $row = $table->Rows->Item($rownum);# code crashes here

...

}
}
[download]

Win32::OLE(0.1712) error 0x8002000e: "Invalid number of parameters" in METHOD/PROPERTYGET "Item" at C:/Strawberry/perl/lib/Devel/ptkdb.pm line 2658.

Greeting Monks,

I’m experimenting with using the built-in `DB` package to build a lightweight call tracer. I’ve previously done something similar using **Hook::LexWrap** and **Module::Info**, but this time I’d like to catch **entry and exit** events for subroutines directly through Perl’s built-in debugging interface.

Here’s the minimal prototype I’ve tried:

                                                                      
+        
package Devel::Lite;

use strict;
use warnings;
use Time::HiRes ();

our $IN_TRACE = 0;

sub DB::sub {
    return &$DB::sub(@_) if $IN_TRACE;
    local $IN_TRACE = 1;

    my $sub = $DB::sub;
    my @args = @_;

    print STDERR ">> entering $sub\n";
    my @ret;
    eval { @ret = &$sub(@args); };
    print STDERR "<< leaving $sub\n";

    die $@ if $@;
    return wantarray ? @ret : $ret[0];
}
1;
[download]

Then I run it like this:

perl -I. -d:TobyLite -e "sub x { print qq{okay\n}; } x();"
[download]

…but it loops endlessly and never reaches my test sub.

If I insert a print inside the loop, I see repeated `"Time::HiRes::time"`, suggesting recursion into the debugger hook itself.

I’ve tried things like skipping `Devel::Lite` and `Time::HiRes` calls explicitly, and guarding with `$IN_TRACE`, but the loop persists.

What I’d like to understand is:

1. Why is `DB::sub` being triggered recursively despite the guard?

2. What’s the correct minimal pattern for tracing sub entry/exit safely?

3. Bonus: how best to distinguish user code from CPAN/core subs to keep the trace clean?

Environment:

This is perl 5.38.x built for MSWin32-x64-multi-thread

Running on Windows 10.

Any insight or examples would be appreciated — I’m not trying to build a full debugger, just a simple tracer that logs the call tree.

TIA

PS: So far this has stumped (in no order) ChatGPT, claude, and Gemini.