From 42ff67d458b2682d18f57843eb73e94600511dfb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Stefan=20Br=C3=BCns?= <stefan.bruens@rwth-aachen.de>
Date: Thu, 17 May 2018 02:35:32 +0200
Subject: [PATCH] GfxFont: Reject invalid names from encoding table

Some PDF files contain invalid glyph names in the Encoding /Differences
array. Putting these e.g. in a Type42 /Encoding array when converting it
to postscript creates an invalid file, rejected by e.g. ghostscript.

If the encoding already has a name for a codepoint we keep it, otherwise
(i.e. the name is empty or .notdef) we create a name.
---
 poppler/GfxFont.cc | 32 +++++++++++++++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)

diff --git a/poppler/GfxFont.cc b/poppler/GfxFont.cc
index 7b4d4b54..b9f150f3 100644
--- a/poppler/GfxFont.cc
+++ b/poppler/GfxFont.cc
@@ -944,6 +944,25 @@ static GBool testForNumericNames(Dict *fontDict, GBool hex) {
   return numeric;
 }
 
+// Return gTrue if the name is valid, according to
+// https://github.com/adobe-type-tools/agl-specification
+static GBool glyphNameIsValid(const char *name)
+{
+  // Valid characters: [a-zA-Z0-9_.]
+  // The AGL spec has more requirements, but this breaks some documents
+  while (*name != '\0') {
+    if (islower(*name) || isdigit(*name) ||
+        isupper(*name) || (*name == '_') ||
+        (*name == '.')) {
+      name++;
+    } else {
+      return gFalse;
+    }
+  }
+
+  return gTrue;
+}
+
 Gfx8BitFont::Gfx8BitFont(XRef *xref, const char *tagA, Ref idA, GooString *nameA,
 			 GfxFontType typeA, Ref embFontIDA, Dict *fontDict):
   GfxFont(tagA, idA, nameA, typeA, embFontIDA) {
@@ -1221,12 +1240,23 @@ Gfx8BitFont::Gfx8BitFont(XRef *xref, const char *tagA, Ref idA, GooString *nameA
 	if (obj3.isInt()) {
 	  code = obj3.getInt();
 	} else if (obj3.isName()) {
-	  if (code >= 0 && code < 256) {
+	  if (glyphNameIsValid(obj3.getName()) && code >= 0 && code < 256) {
 	    if (encFree[code]) {
 	      gfree(enc[code]);
 	    }
 	    enc[code] = copyString(obj3.getName());
 	    encFree[code] = gTrue;
+	  } else if (code >= 0 && code < 256) {
+	    // if there is no name yet, we create one
+	    if (!enc[code] || !strcmp(enc[code], ".notdef")) {
+	      if (encFree[code]) {
+		gfree(enc[code]);
+	      }
+	      char namebuf[8] = "uni0000";
+	      snprintf(namebuf, 8, "uni%04x", code);
+	      enc[code] = copyString(namebuf);
+	      encFree[code] = gTrue;
+	    }
 	  }
 	  ++code;
 	} else {
-- 
2.16.3