From aa730216358403d7a94ff54195de9f4dfd27f402 Mon Sep 17 00:00:00 2001
From: Reinhold Kainhofer <reinhold@kainhofer.com>
Date: Tue, 26 Apr 2011 12:11:33 +0200
Subject: [PATCH] Fix "make doc" error: Character set messup in pdf-scheme.cc

In "make doc", somehow the LANG env variable is cleared, so
glib returned charset=ANSI_X3.4-1968 for the input string. This would
disallow any accented characters in the input string.
However, we know that the input string is by definition always UTF-8,
so the solution is not to use the current locale, but hardcode UTF-8.

Once we switch to guile 2.0, we should use (sting->utf16 str 'big)
instead of manually converting the string in C.
---
 lily/pdf-scheme.cc   | 63 ++++++++++++++++++++++++++++++--------------
 scm/framework-ps.scm |  2 ++
 2 files changed, 45 insertions(+), 20 deletions(-)

diff --git a/lily/pdf-scheme.cc b/lily/pdf-scheme.cc
index 6d717c55ad..d9b7aec601 100644
--- a/lily/pdf-scheme.cc
+++ b/lily/pdf-scheme.cc
@@ -20,41 +20,64 @@
 #include <glib.h>
 using namespace std;
 
+#include "international.hh"
+#include "warn.hh"
 #include "lily-guile.hh"
 
 
 LY_DEFINE (ly_encode_string_for_pdf, "ly:encode-string-for-pdf",
 	   1, 0, 0, (SCM str),
-	   "Check whether the string needs to be encoded for PDF output (Latin1,"
-	   " PDFDocEncoding or in the most general case UTF-16BE).")
+	   "Encode the given string to either Latin1 (which is a subset of"
+	   " the PDFDocEncoding) or if that's not possible to full UTF-16BE"
+	   " with Byte-Order-Mark (BOM).")
 {
   LY_ASSERT_TYPE (scm_is_string, str, 1);
   char *p = ly_scm2str0 (str);
   char *g = NULL;
-  const char *charset;
+  char const *charset="ISO-8859-1"; // Input is ALWAYS UTF-8!
   gsize bytes_written = 0;
-  g_get_charset (&charset); /* The current locale */
 
-  /* First, try to convert to ISO-8859-1 (no encodings required) */
+  /* First, try to convert to ISO-8859-1 (no encodings required). This will
+   * fail, if the string contains accented characters, so we do not check
+   * for errors. */
   g = g_convert (p, -1, "ISO-8859-1", charset, 0, &bytes_written, 0);
   /* If that fails, we have to resolve to full UTF-16BE */
-  if (!g) {
-    char *g_without_BOM = g_convert (p, -1,  "UTF-16BE", charset, 0, &bytes_written, 0);
-    /* prepend the BOM manually, g_convert doesn't do it! */
-    g = new char[bytes_written+3];
-    g[0] = (char)254;
-    g[1] = (char)255;
-    memcpy (&g[2], g_without_BOM, bytes_written+1); // Copy string + \0
-    free (g_without_BOM);
-    bytes_written += 2;
-  }
+  if (!g) 
+    {
+      GError *e = NULL;
+      char *g_without_BOM = g_convert (p, -1,  "UTF-16BE", charset, 0, &bytes_written, &e);
+      if (e != NULL) 
+        {
+          warning (_f("Conversion of string `%s' to UTF-16be failed: %s", p, e->message));
+          g_error_free (e);
+        }
+      /* UTF-16BE allows/recommends a byte-order-mark (BOM) of two bytes 
+       * \xFE\xFF at the begin of the string. The pdfmark specification 
+       * requires it and depends on it to distinguish PdfDocEncoding from
+       * UTF-16BE. As g_convert does not automatically prepend this BOM 
+       * for UTF-16BE (only for UTF-16, which uses lower endian by default, 
+       * though), we have to prepend it manually. */
+      if (g_without_BOM) // conversion to UTF-16be might have failed (shouldn't!)
+        { 
+          g = new char[bytes_written+3];
+          char const *BOM = "\xFE\xFF";
+          strcpy (g, BOM);
+          memcpy (&g[2], g_without_BOM, bytes_written+1); // Copy string + \0
+          free (g_without_BOM);
+          bytes_written += 2;
+        }
+    }
   free (p);
 
   /* Convert back to SCM object and return it */
-  if (g) {
-    return scm_from_locale_stringn (g, bytes_written);
-  } else {
+  /* FIXME guile-2.0: With guile 2.0 the internal representation of a string
+   *                  has changed (char vector rather than binary bytes in 
+   *                  UTF-8). However, with guile 2.0, ly:encode-string-for-pdf
+   *                  is no longer needed and can be replaced by the new 
+   *                  (string->utf16 str 'big)
+   */
+  if (g) 
+    return scm_take_str (g, bytes_written); // scm_take_str eventually frees g!
+  else
     return str;
-  }
-
 }
diff --git a/scm/framework-ps.scm b/scm/framework-ps.scm
index b0359bcd88..7c44920d67 100644
--- a/scm/framework-ps.scm
+++ b/scm/framework-ps.scm
@@ -416,6 +416,8 @@
   (define (metadata-encode val)
     ;; First, call ly:encode-string-for-pdf to encode the string (latin1 or
     ;; utf-16be), then escape all parentheses and backslashes
+    ;; FIXME guile-2.0: use (string->utf16 str 'big) instead
+
     (ps-quote (ly:encode-string-for-pdf val)))
   (define (metadata-lookup-output overridevar fallbackvar field)
     (let* ((overrideval (ly:modules-lookup (list header) overridevar))
-- 
2.39.2