[PATCH 4/3 v2 (bugfix)] gitweb: Fix fallback mode of to_utf8 subroutine

e5d3de5 (gitweb: use Perl built-in utf8 function for UTF-8 decoding.,
2007-12-04) was meant to make gitweb faster by using Perl's internals
(see subsection "Messing with Perl's Internals" in Encode(3pm) manpage)

Simple benchmark confirms that (old = 00f429a, new = this version):
note that it is synthetic benchmark of standalone subroutines, not
of gitweb itself (where probably no visible difference in performace
will show)

        Rate  old  new
  old 1582/s   -- -64%
  new 4453/s 181%   --

Unfortunately it made fallback mode of to_utf8 do not work...  except
for default value 'latin1' of $fallback_encoding (because 'latin1' is
Perl native encoding), which is probably why it was not noticed for so
long.


utf8::valid(STRING) is an internal function that tests whether STRING
is in a _consistent state_ regarding UTF-8.  It returns true is
well-formed UTF-8 and has the UTF-8 flag on _*or*_ if string is held
as bytes (both these states are 'consistent').

For gitweb in most cases the second option was true, as output from
git commands is opened without ':utf8' layer.  So utf8::valid is not
useful for to_utf8.

What made it look as if to_utf8() fallback mode worked correctly
(though only for $fallback_encoding at its default value 'latin1')
was the fact that utf8::decode(STRING) turns on UTF-8 flag only if
source string^W octets form a valid UTF-8 and it contains multi-byte
UTF-8 characters... this means that if string was not valid UTF-8
it didn't get UTF-8 flag.

When string doesn't have UTF-8 flag set, it is treated as if it was in
native Perl encoding, which is 'latin1' (unless native encoding is
EBCDIC ;-)).  So it was ':utf8' layer that actually converted 'latin1'
(no UTF-8 flag == native == 'latin1) to 'utf8', and not to_utf8()
subroutine.  Fallback mode was never triggered.


Let's make use of the fact that utf8::decode(STRING) returns false if
STRING is invalid as UTF-8 to check whether to enable fallback mode.

Note however that if STRING has UTF-8 flag set already, then
utf8::decode also returns false, which could cause problems if given
string was already converted with to_utf8().  Such double conversion
can happen in gitweb.  Therefore we have to check if STRING has UTF-8
flag set with utf8::is_utf8(); if this subroutine returns true then we
have already decoded (converted) string, and don't have to do it
second time.

Signed-off-by: Jakub Narebski <redacted>
---
Jakub Narebski wrote:

P.S. I started to get strange errors

 XML Parsing Error: xml processing instruction not at start of external entity
 Location: http://localhost/cgi-bin/gitweb/gitweb.cgi
 Line Number 37, Column 1:
 <?xml version="1.0" encoding="utf-8"?>
 ^

while "show source" shows that '<?xml version="1.0" encoding="utf-8"?>'
is the first line.  WTF?!?

P.P.S. Now I am getting errors when running gitweb, but only in some
cases (via mod_cgi not as standalone script, only when using lynx),
namely it looks like it falls back to 'latin1' when doing content
which is valid UTF-8.

Will investigate.

Now it is fixed; the error was caused by to_utf8 not dealing with double
encoding for strings outside 7bit ASCII.

 gitweb/gitweb.perl |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl
index d24763b..fc41b07 100755
--- a/gitweb/gitweb.perl
+++ b/gitweb/gitweb.perl

@@ -1443,8 +1443,8 @@ sub validate_refname {
 sub to_utf8 {
 	my $str = shift;
 	return undef unless defined $str;
-	if (utf8::valid($str)) {
-		utf8::decode($str);
+
+	if (utf8::is_utf8($str) || utf8::decode($str)) {
 		return $str;
 	} else {
 		return decode($fallback_encoding, $str, Encode::FB_DEFAULT);

-- 
1.7.6