Merge pull request matplotlib#20400 from QuLogic/stricter-psfontsmap

Make pdftex.map parsing stricter

Author: jklymak

lib/matplotlib/dviread.py

@@ -828,14 +828,19 @@ def __new__(cls, filename):
         # store the unparsed lines (keyed by the first word, which is the
         # texname) and parse them on-demand.
         with open(filename, 'rb') as file:
-            self._unparsed = {line.split(b' ', 1)[0]: line for line in file}
+            self._unparsed = {}
+            for line in file:
+                tfmname = line.split(b' ', 1)[0]
+                self._unparsed.setdefault(tfmname, []).append(line)
         self._parsed = {}
         return self
 
     def __getitem__(self, texname):
         assert isinstance(texname, bytes)
         if texname in self._unparsed:
-            self._parse_and_cache_line(self._unparsed.pop(texname))
+            for line in self._unparsed.pop(texname):
+                if self._parse_and_cache_line(line):
+                    break
         try:
             return self._parsed[texname]
         except KeyError:
@@ -879,6 +884,7 @@ def _parse_and_cache_line(self, line):
         if not line or line.startswith((b" ", b"%", b"*", b";", b"#")):
             return
         tfmname = basename = special = encodingfile = fontfile = None
+        is_subsetted = is_t1 = is_truetype = False
         matches = re.finditer(br'"([^"]*)(?:"|$)|(\S+)', line)
         for match in matches:
             quoted, unquoted = match.groups()
@@ -897,14 +903,13 @@ def _parse_and_cache_line(self, line):
                         encodingfile = word
                     else:
                         fontfile = word
+                        is_subsetted = True
                 elif tfmname is None:
                     tfmname = unquoted
                 elif basename is None:
                     basename = unquoted
             elif quoted:
                 special = quoted
-        if basename is None:
-            basename = tfmname
         effects = {}
         if special:
             words = reversed(special.split())
@@ -913,13 +918,35 @@ def _parse_and_cache_line(self, line):
                     effects["slant"] = float(next(words))
                 elif word == b"ExtendFont":
                     effects["extend"] = float(next(words))
-        if encodingfile is not None and not encodingfile.startswith(b"/"):
+
+        # Verify some properties of the line that would cause it to be ignored
+        # otherwise.
+        if fontfile is not None:
+            if fontfile.endswith((b".ttf", b".ttc")):
+                is_truetype = True
+            elif not fontfile.endswith(b".otf"):
+                is_t1 = True
+        elif basename is not None:
+            is_t1 = True
+        if is_truetype and is_subsetted and encodingfile is None:
+            return
+        if not is_t1 and ("slant" in effects or "extend" in effects):
+            return
+        if abs(effects.get("slant", 0)) > 1:
+            return
+        if abs(effects.get("extend", 0)) > 2:
+            return
+
+        if basename is None:
+            basename = tfmname
+        if encodingfile is not None:
             encodingfile = find_tex_file(encodingfile)
-        if fontfile is not None and not fontfile.startswith(b"/"):
+        if fontfile is not None:
             fontfile = find_tex_file(fontfile)
         self._parsed[tfmname] = PsFont(
             texname=tfmname, psname=basename, effects=effects,
             encoding=encodingfile, filename=fontfile)
+        return True
 
 
 def _parse_enc(path):

lib/matplotlib/tests/baseline_images/dviread/test.map

@@ -2,9 +2,32 @@
 TeXfont1 PSfont1 <font1.pfb <font1.enc
 TeXfont2 PSfont2 <font2.enc <font2.pfa
 TeXfont3 PSfont3 "1.23 UnknownEffect" <[enc3.foo < font3.pfa
-TeXfont4 PSfont4 "-0.1 SlantFont 2.2 ExtendFont" <font4.enc <font4.pfa
+TeXfont4 PSfont4 "-0.1 SlantFont 1.2 ExtendFont" <font4.enc <font4.pfa
 TeXfont5 PSfont5 <encoding1.enc <encoding2.enc <font5.pfb
 TeXfont6 PSfont6
 TeXfont7 PSfont7 < font7.enc
 TeXfont8 PSfont8 <<font8.pfb
 TeXfont9 </absolute/font9.pfb
+% Only the first of a duplicate key is used.
+TeXfontA PSfontA1
+TeXfontA PSfontA2
+%
+% Check SlantFont/ExtendFont
+%
+% Options are only allowed on T1 fonts (3 TrueType, 1 bitmap, and 1 T1 below).
+TeXfontB PSfontB1 "-0.1 SlantFont 1.2 ExtendFont" <fontB1.enc <fontB1.ttf
+TeXfontB PSfontB2 "-0.1 SlantFont" <fontB2.enc <fontB2.ttc
+TeXfontB PSfontB3 "1.2 ExtendFont" <fontB3.enc <fontB3.otf
+TeXfontB "0.1 SlantFont 1.2 ExtendFont"
+% Also, only within range ±1 / ±2.
+TeXfontB PSfontB4 "1.1 SlantFont 1.2 ExtendFont"
+TeXfontB PSfontB5 "0.1 SlantFont 2.2 ExtendFont"
+% This is the only one that works:
+TeXfontB PSfontB6
+%
+% Invalid TrueType entry.
+%
+% Must have encodingfile if subsetted.
+TeXfontC PSfontC1 <fontC1.ttf
+TeXfontC PSfontC2 <fontC2.ttc
+TeXfontC PSfontC3 <fontC3.ttf <8r.enc

lib/matplotlib/tests/test_dviread.py

@@ -28,7 +28,7 @@ def test_PsfontsMap(monkeypatch):
         else:
             assert entry.filename == b'font%d.pfb' % n
         if n == 4:
-            assert entry.effects == {'slant': -0.1, 'extend': 2.2}
+            assert entry.effects == {'slant': -0.1, 'extend': 1.2}
         else:
             assert entry.effects == {}
     # Some special cases
@@ -44,6 +44,15 @@ def test_PsfontsMap(monkeypatch):
     entry = fontmap[b'TeXfont9']
     assert entry.psname == b'TeXfont9'
     assert entry.filename == b'/absolute/font9.pfb'
+    # First of duplicates only.
+    entry = fontmap[b'TeXfontA']
+    assert entry.psname == b'PSfontA1'
+    # Slant/Extend only works for T1 fonts.
+    entry = fontmap[b'TeXfontB']
+    assert entry.psname == b'PSfontB6'
+    # Subsetted TrueType must have encoding.
+    entry = fontmap[b'TeXfontC']
+    assert entry.psname == b'PSfontC3'
     # Missing font
     with pytest.raises(KeyError, match='no-such-font'):
         fontmap[b'no-such-font']