diff mbox series

gitweb: fix UTF-8 encoding when using CGI::Fast

Message ID 20200329002028.26080-1-julm+git@sourcephile.fr (mailing list archive)
State New, archived
Headers show
Series gitweb: fix UTF-8 encoding when using CGI::Fast | expand

Commit Message

Julien Moutinho March 29, 2020, 12:20 a.m. UTC
FCGI streams are implemented using the older stream API: TIEHANDLE,
therefore applying PerlIO layers using binmode() has no effect to them.
The solution in this patch is to redefine the FCGI::Stream::PRINT function
to use UTF-8 as output encoding, except within git_blob_plain() and git_snapshot()
which must still output in raw binary mode.

This problem and solution were previously reported back in 2012:
- http://git.661346.n2.nabble.com/Gitweb-running-as-FCGI-does-not-print-its-output-in-UTF-8-td7573415.html
- http://stackoverflow.com/questions/5005104

Signed-off-by: Julien Moutinho <julm+git@sourcephile.fr>
---
 gitweb/gitweb.perl | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

Comments

Junio C Hamano March 29, 2020, 4:06 p.m. UTC | #1
Julien Moutinho <julm+git@sourcephile.fr> writes:

>  	require CGI::Fast;
>  	our $CGI = 'CGI::Fast';
> +	# FCGI is not Unicode aware hence the UTF-8 encoding must be done manually.
> +	# However no encoding must be done within git_blob_plain() and git_snapshot()
> +	# which must still output in raw binary mode.

I guess this comment would be sufficient to help future developers
when they find that newer version of CGI::Fast has become Unicode
aware later can make this part conditional to the version of the
module, perhaps?

Would "use CGI::Fast (-utf8)" instead of the whole thing help, by
the way?

> +	no warnings 'redefine';
> +	my $enc = Encode::find_encoding('UTF-8');
> +	*FCGI::Stream::PRINT = sub {
> +		my @OUTPUT = @_;
> +		for (my $i = 1; $i < @_; $i++) {
> +			$OUTPUT[$i] = $enc->encode($_[$i], Encode::FB_CROAK|Encode::LEAVE_SRC);
> +		}
> +		@_ = @OUTPUT;
> +		goto $FCGI_Stream_PRINT_raw;
> +	};



>  	my $request_number = 0;
>  	# let each child service 100 requests
> @@ -7079,6 +7093,7 @@ sub git_blob_plain {
>  			($sandbox ? 'attachment' : 'inline')
>  			. '; filename="' . $save_as . '"');
>  	local $/ = undef;
> +	local *FCGI::Stream::PRINT = $FCGI_Stream_PRINT_raw;
>  	binmode STDOUT, ':raw';
>  	print <$fd>;
>  	binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi
> @@ -7417,6 +7432,7 @@ sub git_snapshot {
>  
>  	open my $fd, "-|", $cmd
>  		or die_error(500, "Execute git-archive failed");
> +	local *FCGI::Stream::PRINT = $FCGI_Stream_PRINT_raw;
>  	binmode STDOUT, ':raw';
>  	print <$fd>;
>  	binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi
Julien Moutinho March 29, 2020, 7:14 p.m. UTC | #2
Le dim. 29 mars 2020 09h06 -0700, Junio C Hamano a écrit :
> Julien Moutinho <julm+git@sourcephile.fr> writes:
> >  	require CGI::Fast;
> >  	our $CGI = 'CGI::Fast';
> > +	# FCGI is not Unicode aware hence the UTF-8 encoding must be done manually.
> > +	# However no encoding must be done within git_blob_plain() and git_snapshot()
> > +	# which must still output in raw binary mode.
> 
> I guess this comment would be sufficient to help future developers
> when they find that newer version of CGI::Fast has become Unicode
> aware later can make this part conditional to the version of the
> module, perhaps?
Sure, though as long as CGI::Fast will be relying on FCGI,
I would not bet on any improvement on this bug
which has been waiting to be fixed on FCGI's bugtracker since 2013:
https://rt.cpan.org/Public/Bug/Display.html?id=89383

> Would "use CGI::Fast (-utf8)" instead of the whole thing help, by
> the way?
Unfortunately not, the -utf8 option (aka. $CGI::$PARAM_UTF8)
controls the decoding of the input parameters,
not the encoding of the output.
- https://metacpan.org/pod/CGI#-utf8
- https://stackoverflow.com/questions/5005104/how-to-force-fastcgi-to-encode-form-data-as-utf-8-as-cgi-pm-has-option/7097698#7097698

> > our $FCGI_Stream_PRINT_raw = \&FCGI::Stream::PRINT;
> [...]
> > +	local *FCGI::Stream::PRINT = $FCGI_Stream_PRINT_raw;
I had forgotten to test the patch without FastCGI,
but AFAICS it is innocuous in non-FastCGI mode:
Perl does not chokes on \&FCGI::Stream::PRINT despite it being not reachable,
and the local binding emits no redefine warning since it's not a redefine
in this case, but a define.

Regards,
diff mbox series

Patch

diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl
index 65a3a9e..1a02a12 100755
--- a/gitweb/gitweb.perl
+++ b/gitweb/gitweb.perl
@@ -1291,9 +1291,23 @@  sub run_request {
 our ($pre_dispatch_hook, $post_dispatch_hook, $pre_listen_hook);
 our $CGI = 'CGI';
 our $cgi;
+our $FCGI_Stream_PRINT_raw = \&FCGI::Stream::PRINT;
 sub configure_as_fcgi {
 	require CGI::Fast;
 	our $CGI = 'CGI::Fast';
+	# FCGI is not Unicode aware hence the UTF-8 encoding must be done manually.
+	# However no encoding must be done within git_blob_plain() and git_snapshot()
+	# which must still output in raw binary mode.
+	no warnings 'redefine';
+	my $enc = Encode::find_encoding('UTF-8');
+	*FCGI::Stream::PRINT = sub {
+		my @OUTPUT = @_;
+		for (my $i = 1; $i < @_; $i++) {
+			$OUTPUT[$i] = $enc->encode($_[$i], Encode::FB_CROAK|Encode::LEAVE_SRC);
+		}
+		@_ = @OUTPUT;
+		goto $FCGI_Stream_PRINT_raw;
+	};
 
 	my $request_number = 0;
 	# let each child service 100 requests
@@ -7079,6 +7093,7 @@  sub git_blob_plain {
 			($sandbox ? 'attachment' : 'inline')
 			. '; filename="' . $save_as . '"');
 	local $/ = undef;
+	local *FCGI::Stream::PRINT = $FCGI_Stream_PRINT_raw;
 	binmode STDOUT, ':raw';
 	print <$fd>;
 	binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi
@@ -7417,6 +7432,7 @@  sub git_snapshot {
 
 	open my $fd, "-|", $cmd
 		or die_error(500, "Execute git-archive failed");
+	local *FCGI::Stream::PRINT = $FCGI_Stream_PRINT_raw;
 	binmode STDOUT, ':raw';
 	print <$fd>;
 	binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi