Changeset 3073 for trunk/gsdl

Timestamp:

2002-04-03T15:44:42+12:00 (22 years ago)

Author:

jrm21

Message:

1) Default Title now correctly escapes [ and ] chars.
2) "Content-type: text" is now allowed (for emails pre-1996 RFC for all the

people with old email archives).

3) Check if text is really encoded with win1252 when it claims to be ascii,

as some mailers on Microsoft Windows do this, which causes our XML parser
to fail.

File:

• trunk/gsdl/perllib/plugins/EMAILPlug.pm (modified) (3 diffs)

trunk/gsdl/perllib/plugins/EMAILPlug.pm

@@ -279 +279 @@
     my $mimeinfo="";
     my $charset = "iso_8859_1";
-    # Do MIME and encoding stuff
-    if ($Headers =~ /^content\-type:\s*([\w\/\-]+)\s*\;?\s*(.+?)\s*$/mi)
+    # Do MIME and encoding stuff. Allow \s in mimeinfo in case there is
+    # more than one parameter given to Content-type.
+    # eg: Content-type: text/plain; charset="us-ascii"; format="flowed"
+    if ($Headers =~ /^content\-type:\s*([\w\/\-]+)\s*(\;\s*.*)\s*$/mi)
     {
         $mimetype=$1;
         $mimetype =~ tr/[A-Z]/[a-z]/;
+
+        if ($mimetype eq "text") { # for pre-RFC2045 messages (c. 1996)
+        $mimetype = "text/plain";
+        }
+
         $mimeinfo=$2;
-        if ($mimeinfo =~ /charset=\"([^\"]+)\"/) {
+        if (!defined $mimeinfo) {
+        $mimeinfo="";
+        } else { # strip leading and trailing stuff
+        $mimeinfo =~ s/^\;\s*//;
+        $mimeinfo =~ s/\s*$//;
+        }
+        if ($mimeinfo =~ /charset=\"([^\"]+)\"/i) {
           $charset = $1;
         }
@@ -325 +338 @@
     $Title .= "<br>From: " . text_into_html($raw{'From'});
     $Title .= "<br>Date: " . text_into_html($raw{'DateText'});
+    $Title =~ s/\[/&#91;/g; $Title =~ s/\]/&#93;/g;
 
     $doc_obj->add_utf8_metadata ($cursection, "Title", $Title);
@@ -760 +774 @@
   $charset = "iso_8859_1" if ($charset eq "us_ascii" || $charset eq "ascii");
 
+  if ($charset eq "iso_8859_1") {
+      # test if the mailer lied, and it has win1252 chars in it...
+      # 1252 has characters between 0x80 and 0x9f, 8859-1 doesn't
+      if ($$textref =~ m/[\x80-\x9f]/) {
+      my $outhandle = $self->{'outhandle'};
+      print $outhandle "EMAILPlug: Headers claim ISO charset but MS ";
+      print $outhandle "codepage 1252 detected.\n";
+      $charset = "windows_1252";
+      }
+  }
   $$textref=&unicode::unicode2utf8(&unicode::convert2unicode($charset,$textref));
 }