Add Makefile.PL prompt to enable no_network by default

LibXML.pm

@@ -262,6 +262,9 @@ use constant {
 };
 
 our $XML_LIBXML_PARSE_DEFAULTS = ( XML_PARSE_NODICT );
+if ( XML::LibXML::DEFAULT_NONET() ) {
+    $XML_LIBXML_PARSE_DEFAULTS |= XML_PARSE_NONET;
+}
 
 # this hash is made global so that applications can add names for new
 # libxml2 parser flags as temporary workaround
@@ -1088,7 +1091,7 @@ sub _html_options {
   $flags |=    64 if $opts->{suppress_warnings};
   $flags |=   128 if exists $opts->{pedantic_parser} ? $opts->{pedantic_parser} : $self->pedantic_parser;
   $flags |=   256 if exists $opts->{no_blanks} ? $opts->{no_blanks} : !$self->keep_blanks;
-  $flags |=  2048 if exists $opts->{no_network} ? $opts->{no_network} : !$self->no_network;
+  $flags |=  2048 if exists $opts->{no_network} ? $opts->{no_network} : 1;
   $flags |= 16384 if $opts->{no_cdata};
   $flags |= 65536 if $opts->{compact}; # compact small text nodes; no modification
                                          # of the tree allowed afterwards

LibXML.xs

@@ -1661,6 +1661,19 @@ HAVE_READER()
     OUTPUT:
         RETVAL
 
+# Whether Makefile.PL was run with -DLIBXML_DEFAULT_NONET (user chose
+# secure defaults at build time).
+int
+DEFAULT_NONET()
+    CODE:
+#ifdef LIBXML_DEFAULT_NONET
+        RETVAL = 1;
+#else
+        RETVAL = 0;
+#endif
+    OUTPUT:
+        RETVAL
+
 int
 HAVE_THREAD_SUPPORT()
     CODE:

Makefile.PL

@@ -75,6 +75,27 @@ my %xsbuild_concat = (
 if ($ENV{DEVELOPER}) {
   $xsbuild_concat{DEFINE} .= ' -Wall -Wextra -Wno-unused-parameter -Werror=implicit-function-declaration';
 }
+
+# Offer to compile with no_network => 1 as the default for all parsers.
+# Env var lets automated builds skip the interactive prompt.
+if (defined $ENV{XML_LIBXML_DEFAULT_NONET}) {
+  if ($ENV{XML_LIBXML_DEFAULT_NONET}) {
+    $xsbuild_concat{DEFINE} .= ' -DLIBXML_DEFAULT_NONET';
+    print "XML_LIBXML_DEFAULT_NONET=$ENV{XML_LIBXML_DEFAULT_NONET}: enabling no_network by default\n";
+  }
+} else {
+  my $ans = prompt(
+    "\nXML::LibXML can forbid network access during XML parsing by default\n"
+    . "(DTD fetches, entity resolution, schema imports).  This is more secure\n"
+    . "but MAY BREAK existing code that expects network access to be on.\n"
+    . "Enable no_network by default?",
+    "n"
+  );
+  if (lc($ans) =~ /^y/) {
+    $xsbuild_concat{DEFINE} .= ' -DLIBXML_DEFAULT_NONET';
+  }
+}
+
 my %xsbuild = Alien::Base::Wrapper->mm_args;  # Might contain a definition of DEFINE, must thus concatenate.
 while (my ($k, $v) = each %xsbuild_concat) {
   my $base_val = $xsbuild{$k};
@@ -99,7 +120,7 @@ my %WriteMakefileArgs = (
   "MIN_PERL_VERSION" => '5.008001',
   "VERSION_FROM" => "LibXML.pm",
   'META_MERGE' => {
-    'dynamic_config' => 0,
+    'dynamic_config' => 1,
     'meta-spec' => {version => 2},
     'no_index' => {directory => ['xt']},
     'resources' => {

docs/libxml.dbk

@@ -1830,7 +1830,13 @@ $parser->input_callbacks($c);
             <term>no_network</term>
             <listitem>
 	      <para>/parser, html, reader/</para>
-              <para>Forbid network access; possible values are 0 and 1</para>
+              <para>Forbid network access; possible values are 0 and 1.
+                Defaults to 0 unless XML::LibXML was built with
+                <literal>-DLIBXML_DEFAULT_NONET</literal> (the
+                <literal>perl Makefile.PL</literal> prompt "Enable no_network
+                by default?" was answered yes, or
+                <literal>XML_LIBXML_DEFAULT_NONET=1</literal> was set in the
+                environment), in which case it defaults to 1.</para>
               <para>If set to true, all
                 attempts to fetch non-local resources (such as
                 DTD or external entities) will fail (unless
@@ -5801,8 +5807,9 @@ $rngschema = XML::LibXML::RelaxNG-&gt;new( DOM =&gt; $doc, no_network =&gt; 1 );
                     </funcsynopsis>
 
                     <para>The constructor of XML::LibXML::RelaxNG needs to be called with list of parameters. At least location, string or DOM parameter is required to
-                    specify source of schema. Optional parameter no_network set to 1 cause that parser would not access network and optional parameter recover
-                    set 1 cause that parser would not call die() on errors.</para>
+                    specify source of schema. Optional parameter no_network set to 1 causes the parser to forbid network access (this is already the default
+                    when XML::LibXML was built with <literal>LIBXML_DEFAULT_NONET</literal>). Optional parameter recover
+                    set to 1 causes the parser to not call die() on errors.</para>
 
                     <para>It is important, that each schema only have a single source.</para>
 
@@ -5868,8 +5875,9 @@ $xmlschema = XML::LibXML::Schema-&gt;new( string =&gt; $xmlschemastring, no_netw
                     </funcsynopsis>
 
                     <para>The constructor of XML::LibXML::Schema needs to be called with list of parameters. At least location or string parameter is required to
-                    specify source of schema. Optional parameter no_network set to 1 cause that parser would not access network and optional parameter recover
-                    set 1 cause that parser would not call die() on errors.</para>
+                    specify source of schema. Optional parameter no_network set to 1 causes the parser to forbid network access (this is already the default
+                    when XML::LibXML was built with <literal>LIBXML_DEFAULT_NONET</literal>). Optional parameter recover
+                    set to 1 causes the parser to not call die() on errors.</para>
 
                     <para>It is important, that each schema only have a single source.</para>
 

lib/XML/LibXML/Parser.pod

@@ -798,7 +798,8 @@ do not generate XINCLUDE START/END nodes; possible values are 0 and 1
 
 /parser, html, reader/
 
-Forbid network access; possible values are 0 and 1
+Forbid network access; possible values are 0 and 1. Defaults to 0 unless
+XML::LibXML was built with C<<<<<< -DLIBXML_DEFAULT_NONET >>>>>> (the C<<<<<< perl Makefile.PL >>>>>> prompt "Enable no_network by default?" was answered yes, or C<<<<<< XML_LIBXML_DEFAULT_NONET=1 >>>>>> was set in the environment), in which case it defaults to 1.
 
 If set to true, all attempts to fetch non-local resources (such as DTD or
 external entities) will fail (unless custom callbacks are defined).

lib/XML/LibXML/RelaxNG.pod

@@ -33,9 +33,10 @@ validation.
 
 The constructor of XML::LibXML::RelaxNG needs to be called with list of
 parameters. At least location, string or DOM parameter is required to specify
-source of schema. Optional parameter no_network set to 1 cause that parser
-would not access network and optional parameter recover set 1 cause that parser
-would not call die() on errors.
+source of schema. Optional parameter no_network set to 1 causes the parser to
+forbid network access (this is already the default when XML::LibXML was built
+with C<<<<<< LIBXML_DEFAULT_NONET >>>>>>). Optional parameter recover set to 1 causes the parser to not call die() on
+errors.
 
 It is important, that each schema only have a single source.
 

lib/XML/LibXML/Schema.pod

@@ -32,9 +32,9 @@ validation. As of 2.6.32, libxml2 only supports decimal types up to 24 digits
 
 The constructor of XML::LibXML::Schema needs to be called with list of
 parameters. At least location or string parameter is required to specify source
-of schema. Optional parameter no_network set to 1 cause that parser would not
-access network and optional parameter recover set 1 cause that parser would not
-call die() on errors.
+of schema. Optional parameter no_network set to 1 causes the parser to forbid
+network access (this is already the default when XML::LibXML was built with C<<<<<< LIBXML_DEFAULT_NONET >>>>>>). Optional parameter recover set to 1 causes the parser to not call die() on
+errors.
 
 It is important, that each schema only have a single source.
 

t/43options.t

@@ -48,9 +48,10 @@ no_network
 
 
 {
+  my $nonet_default = XML::LibXML::DEFAULT_NONET();
   my $p = XML::LibXML->new();
   for my $opt (@all) {
-    my $ret = 0;
+    my $ret = ($opt eq 'no_network' && $nonet_default) ? 1 : 0;
     # TEST*$all
     ok(
         ($p->get_option($opt)||0) == $ret