From 5d0b00589c7c61b1eae2f4670b50fa8defea1a70 Mon Sep 17 00:00:00 2001
From: "Julien \"_FrnchFrgg_\" RIVAUD" <frnchfrgg@free.fr>
Date: Sun, 3 Sep 2017 20:39:02 +0200
Subject: [PATCH] Fix the Unicode workaround in wscript
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The previous code was supposed to ensure VERSION and PROGRAM_VERSION
were *not* Unicode objects since though they are the correct type to
represent strings in Python 2, too many libraries have strange issues
dealing with them, and UnicodeErrors can creep in with implicit
conversions from/to Unicode objects.

But in fact it did exactly the contrary since str.decode() always
returns Unicode objects, whose type corresponds to the str class in
Python 3. Fix it so that in both Python 2 and 3 the constants are
eventually instances of str.
---
 wscript | 31 ++++++++++++++++++++++++-------
 1 file changed, 24 insertions(+), 7 deletions(-)

diff --git a/wscript b/wscript
index c7e4245024..4b5acb0ad7 100644
--- a/wscript
+++ b/wscript
@@ -157,7 +157,7 @@ def fetch_tarball_revision ():
     if not os.path.exists ('libs/ardour/revision.cc'):
         print ('This tarball was not created correctly - it is missing libs/ardour/revision.cc')
         sys.exit (1)
-    with open('libs/ardour/revision.cc') as f:
+    with open('libs/ardour/revision.cc', 'rb') as f:
         content = f.readlines()
         remove_punctuation_map = dict((ord(char), None) for char in '";')
         return content[1].decode('utf-8').strip().split(' ')[7].translate (remove_punctuation_map)
@@ -182,12 +182,29 @@ else:
     MICRO = '0'
 
 V = MAJOR + '.' + MINOR + '.' + MICRO
-# Ensure that these are not unicode, which
-# can cause odd problems elsewhere. Note that
-# in python3, encode and decode do not return
-# strings, so we have to force the type.
-VERSION = V.encode ('ascii', 'ignore').decode ("utf-8")
-PROGRAM_VERSION = MAJOR.encode ('ascii', 'ignore').decode ("utf-8")
+
+def sanitize(s):
+    # round-trip to remove anything in the string that is not encodable in
+    # ASCII, yet still keep a real (utf8-encoded internally) string.
+    s = s.encode ('ascii', 'ignore').decode ("utf-8")
+    # In Python3, bytes is the class of binary content and encode() returns
+    # bytes to transform a string according to a text encoding; str is the
+    # class of normal strings (utf8-encoded internally) and decode() returns
+    # that type.
+    # Python 2 did not initially cater for encoding problems and can use str
+    # for both binary content and for (decoded) strings. The Unicode type was
+    # added to correspond to Python 3 str, and the Python 2 str type should
+    # only correspond to bytes. Alas, almost everything in the Python 2
+    # ecosystem has been written with str in mind and doesn't handle Unicode
+    # objects correctly. If Python 2 is in use, s will be a Unicode object and
+    # to avoid strange problems later we convert back to str, but in utf-8
+    # nonetheless.
+    if not isinstance(s, str):
+        s = s.encode("utf-8")
+    return s
+VERSION = sanitize(V)
+PROGRAM_VERSION = sanitize(MAJOR)
+del sanitize
 
 if len (sys.argv) > 1 and sys.argv[1] == 'dist':
         if not 'APPNAME' in os.environ: