Merge branch 'tk/p4-metadata-coding-strategies'

"git p4" updates.

* tk/p4-metadata-coding-strategies:
  git-p4: improve encoding handling to support inconsistent encodings

Author: gitster

Documentation/git-p4.txt

@@ -636,7 +636,42 @@ git-p4.pathEncoding::
 	Git expects paths encoded as UTF-8. Use this config to tell git-p4
 	what encoding Perforce had used for the paths. This encoding is used
 	to transcode the paths to UTF-8. As an example, Perforce on Windows
-	often uses "cp1252" to encode path names.
+	often uses "cp1252" to encode path names. If this option is passed
+	into a p4 clone request, it is persisted in the resulting new git
+	repo.
+
+git-p4.metadataDecodingStrategy::
+	Perforce keeps the encoding of a changelist descriptions and user
+	full names as stored by the client on a given OS. The p4v client
+	uses the OS-local encoding, and so different users can end up storing
+	different changelist descriptions or user full names in different
+	encodings, in the same depot.
+	Git tolerates inconsistent/incorrect encodings in commit messages
+	and author names, but expects them to be specified in utf-8.
+	git-p4 can use three different decoding strategies in handling the
+	encoding uncertainty in Perforce: 'passthrough' simply passes the
+	original bytes through from Perforce to git, creating usable but
+	incorrectly-encoded data when the Perforce data is encoded as
+	anything other than utf-8. 'strict' expects the Perforce data to be
+	encoded as utf-8, and fails to import when this is not true.
+	'fallback' attempts to interpret the data as utf-8, and otherwise
+	falls back to using a secondary encoding - by default the common
+	windows encoding 'cp-1252' - with upper-range bytes escaped if
+	decoding with the fallback encoding also fails.
+	Under python2 the default strategy is 'passthrough' for historical
+	reasons, and under python3 the default is 'fallback'.
+	When 'strict' is selected and decoding fails, the error message will
+	propose changing this config parameter as a workaround. If this
+	option is passed into a p4 clone request, it is persisted into the
+	resulting new git repo.
+
+git-p4.metadataFallbackEncoding::
+	Specify the fallback encoding to use when decoding Perforce author
+	names and changelists descriptions using the 'fallback' strategy
+	(see git-p4.metadataDecodingStrategy). The fallback encoding will
+	only be used when decoding as utf-8 fails. This option defaults to
+	cp1252, a common windows encoding. If this option is passed into a
+	p4 clone request, it is persisted into the resulting new git repo.
 
 git-p4.largeFileSystem::
 	Specify the system that is used for large (binary) files. Please note

git-p4.py

@@ -31,6 +31,7 @@
 # pylint: disable=wrong-import-position
 #
 
+import struct
 import sys
 if sys.version_info.major < 3 and sys.version_info.minor < 7:
     sys.stderr.write("git-p4: requires Python 2.7 or later.\n")
@@ -71,6 +72,9 @@
 # The block size is reduced automatically if required
 defaultBlockSize = 1 << 20
 
+defaultMetadataDecodingStrategy = 'passthrough' if sys.version_info.major == 2 else 'fallback'
+defaultFallbackMetadataEncoding = 'cp1252'
+
 p4_access_checked = False
 
 re_ko_keywords = re.compile(br'\$(Id|Header)(:[^$\n]+)?\$')
@@ -229,6 +233,70 @@ def encode_text_stream(s):
         return s.encode('utf_8') if isinstance(s, unicode) else s
 
 
+class MetadataDecodingException(Exception):
+    def __init__(self, input_string):
+        self.input_string = input_string
+
+    def __str__(self):
+        return """Decoding perforce metadata failed!
+The failing string was:
+---
+{}
+---
+Consider setting the git-p4.metadataDecodingStrategy config option to
+'fallback', to allow metadata to be decoded using a fallback encoding,
+defaulting to cp1252.""".format(self.input_string)
+
+
+encoding_fallback_warning_issued = False
+encoding_escape_warning_issued = False
+def metadata_stream_to_writable_bytes(s):
+    encodingStrategy = gitConfig('git-p4.metadataDecodingStrategy') or defaultMetadataDecodingStrategy
+    fallbackEncoding = gitConfig('git-p4.metadataFallbackEncoding') or defaultFallbackMetadataEncoding
+    if not isinstance(s, bytes):
+        return s.encode('utf_8')
+    if encodingStrategy == 'passthrough':
+        return s
+    try:
+        s.decode('utf_8')
+        return s
+    except UnicodeDecodeError:
+        if encodingStrategy == 'fallback' and fallbackEncoding:
+            global encoding_fallback_warning_issued
+            global encoding_escape_warning_issued
+            try:
+                if not encoding_fallback_warning_issued:
+                    print("\nCould not decode value as utf-8; using configured fallback encoding %s: %s" % (fallbackEncoding, s))
+                    print("\n(this warning is only displayed once during an import)")
+                    encoding_fallback_warning_issued = True
+                return s.decode(fallbackEncoding).encode('utf_8')
+            except Exception as exc:
+                if not encoding_escape_warning_issued:
+                    print("\nCould not decode value with configured fallback encoding %s; escaping bytes over 127: %s" % (fallbackEncoding, s))
+                    print("\n(this warning is only displayed once during an import)")
+                    encoding_escape_warning_issued = True
+                escaped_bytes = b''
+                # bytes and strings work very differently in python2 vs python3...
+                if str is bytes:
+                    for byte in s:
+                        byte_number = struct.unpack('>B', byte)[0]
+                        if byte_number > 127:
+                            escaped_bytes += b'%'
+                            escaped_bytes += hex(byte_number)[2:].upper()
+                        else:
+                            escaped_bytes += byte
+                else:
+                    for byte_number in s:
+                        if byte_number > 127:
+                            escaped_bytes += b'%'
+                            escaped_bytes += hex(byte_number).upper().encode()[2:]
+                        else:
+                            escaped_bytes += bytes([byte_number])
+                return escaped_bytes
+
+        raise MetadataDecodingException(s)
+
+
 def decode_path(path):
     """Decode a given string (bytes or otherwise) using configured path
        encoding options.
@@ -786,11 +854,12 @@ def p4CmdList(cmd, stdin=None, stdin_mode='w+b', cb=None, skip_info=False,
             if bytes is not str:
                 # Decode unmarshalled dict to use str keys and values, except for:
                 #   - `data` which may contain arbitrary binary data
-                #   - `depotFile[0-9]*`, `path`, or `clientFile` which may contain non-UTF8 encoded text
+                #   - `desc` or `FullName` which may contain non-UTF8 encoded text handled below, eagerly converted to bytes
+                #   - `depotFile[0-9]*`, `path`, or `clientFile` which may contain non-UTF8 encoded text, handled by decode_path()
                 decoded_entry = {}
                 for key, value in entry.items():
                     key = key.decode()
-                    if isinstance(value, bytes) and not (key in ('data', 'path', 'clientFile') or key.startswith('depotFile')):
+                    if isinstance(value, bytes) and not (key in ('data', 'desc', 'FullName', 'path', 'clientFile') or key.startswith('depotFile')):
                         value = value.decode()
                     decoded_entry[key] = value
                 # Parse out data if it's an error response
@@ -800,6 +869,10 @@ def p4CmdList(cmd, stdin=None, stdin_mode='w+b', cb=None, skip_info=False,
             if skip_info:
                 if 'code' in entry and entry['code'] == 'info':
                     continue
+            if 'desc' in entry:
+                entry['desc'] = metadata_stream_to_writable_bytes(entry['desc'])
+            if 'FullName' in entry:
+                entry['FullName'] = metadata_stream_to_writable_bytes(entry['FullName'])
             if cb is not None:
                 cb(entry)
             else:
@@ -1603,7 +1676,13 @@ def getUserMapFromPerforceServer(self):
         for output in p4CmdList(["users"]):
             if "User" not in output:
                 continue
-            self.users[output["User"]] = output["FullName"] + " <" + output["Email"] + ">"
+            # "FullName" is bytes. "Email" on the other hand might be bytes
+            # or unicode string depending on whether we are running under
+            # python2 or python3. To support
+            # git-p4.metadataDecodingStrategy=fallback, self.users dict values
+            # are always bytes, ready to be written to git.
+            emailbytes = metadata_stream_to_writable_bytes(output["Email"])
+            self.users[output["User"]] = output["FullName"] + b" <" + emailbytes + b">"
             self.emails[output["Email"]] = output["User"]
 
         mapUserConfigRegex = re.compile(r"^\s*(\S+)\s*=\s*(.+)\s*<(\S+)>\s*$", re.VERBOSE)
@@ -1613,26 +1692,28 @@ def getUserMapFromPerforceServer(self):
                 user = mapUser[0][0]
                 fullname = mapUser[0][1]
                 email = mapUser[0][2]
-                self.users[user] = fullname + " <" + email + ">"
+                fulluser = fullname + " <" + email + ">"
+                self.users[user] = metadata_stream_to_writable_bytes(fulluser)
                 self.emails[email] = user
 
-        s = ''
+        s = b''
         for (key, val) in self.users.items():
-            s += "%s\t%s\n" % (key.expandtabs(1), val.expandtabs(1))
+            keybytes = metadata_stream_to_writable_bytes(key)
+            s += b"%s\t%s\n" % (keybytes.expandtabs(1), val.expandtabs(1))
 
-        open(self.getUserCacheFilename(), 'w').write(s)
+        open(self.getUserCacheFilename(), 'wb').write(s)
         self.userMapFromPerforceServer = True
 
     def loadUserMapFromCache(self):
         self.users = {}
         self.userMapFromPerforceServer = False
         try:
-            cache = open(self.getUserCacheFilename(), 'r')
+            cache = open(self.getUserCacheFilename(), 'rb')
             lines = cache.readlines()
             cache.close()
             for line in lines:
-                entry = line.strip().split("\t")
-                self.users[entry[0]] = entry[1]
+                entry = line.strip().split(b"\t")
+                self.users[entry[0].decode('utf_8')] = entry[1]
         except IOError:
             self.getUserMapFromPerforceServer()
 
@@ -3229,7 +3310,8 @@ def make_email(self, userid):
         if userid in self.users:
             return self.users[userid]
         else:
-            return "%s <a@b>" % userid
+            userid_bytes = metadata_stream_to_writable_bytes(userid)
+            return b"%s <a@b>" % userid_bytes
 
     def streamTag(self, gitStream, labelName, labelDetails, commit, epoch):
         """Stream a p4 tag.
@@ -3253,9 +3335,10 @@ def streamTag(self, gitStream, labelName, labelDetails, commit, epoch):
             email = self.make_email(owner)
         else:
             email = self.make_email(self.p4UserId())
-        tagger = "%s %s %s" % (email, epoch, self.tz)
 
-        gitStream.write("tagger %s\n" % tagger)
+        gitStream.write("tagger ")
+        gitStream.write(email)
+        gitStream.write(" %s %s\n" % (epoch, self.tz))
 
         print("labelDetails=", labelDetails)
         if 'Description' in labelDetails:
@@ -3351,12 +3434,12 @@ def commit(self, details, files, branch, parent="", allow_empty=False):
         self.gitStream.write("commit %s\n" % branch)
         self.gitStream.write("mark :%s\n" % details["change"])
         self.committedChanges.add(int(details["change"]))
-        committer = ""
         if author not in self.users:
             self.getUserMapFromPerforceServer()
-        committer = "%s %s %s" % (self.make_email(author), epoch, self.tz)
 
-        self.gitStream.write("committer %s\n" % committer)
+        self.gitStream.write("committer ")
+        self.gitStream.write(self.make_email(author))
+        self.gitStream.write(" %s %s\n" % (epoch, self.tz))
 
         self.gitStream.write("data <<EOT\n")
         self.gitStream.write(details["desc"])
@@ -4257,6 +4340,14 @@ def run(self, args):
         if self.useClientSpec_from_options:
             system(["git", "config", "--bool", "git-p4.useclientspec", "true"])
 
+        # persist any git-p4 encoding-handling config options passed in for clone:
+        if gitConfig('git-p4.metadataDecodingStrategy'):
+            system(["git", "config", "git-p4.metadataDecodingStrategy", gitConfig('git-p4.metadataDecodingStrategy')])
+        if gitConfig('git-p4.metadataFallbackEncoding'):
+            system(["git", "config", "git-p4.metadataFallbackEncoding", gitConfig('git-p4.metadataFallbackEncoding')])
+        if gitConfig('git-p4.pathEncoding'):
+            system(["git", "config", "git-p4.pathEncoding", gitConfig('git-p4.pathEncoding')])
+
         return True
 
 

t/lib-git-p4.sh

@@ -142,10 +142,11 @@ start_p4d () {
 
 p4_add_user () {
 	name=$1 &&
+	fullname="${2:-Dr. $1}"
 	p4 user -f -i <<-EOF
 	User: $name
 	Email: $name@example.com
-	FullName: Dr. $name
+	FullName: $fullname
 	EOF
 }