From: Arlo Breault <abreault@wikimedia.org>
Date: Wed, 7 Jan 2015 20:03:02 +0000 (-0800)
Subject: T105242: U is for urlencode
X-Git-Tag: 1.31.0-rc.0~10837^2
X-Git-Url: http://git.cyclocoop.org/%24image?a=commitdiff_plain;h=baddcab658195e9f2438b24f15030564e2fbd925;p=lhc%2Fweb%2Fwiklou.git

T105242: U is for urlencode

 * Urlencode and extension tags don't play nice. The current behaviour
   is pretty broken. For example,

      {{urlencode:<pre id="one">two</pre>}}

   outputs

       <pre id="one">two</pre>

   which isn't exactly urlencoded.

 * Just drop the unexpanded content; anything more seems unsafe. There's
   precedent for this already in anchorencode.

 * Noticed in the infobox caption for enwiki/%22F%22_Is_for_Fugitive

Change-Id: I0ec5d2fdeb71f762036464f65029afd029e8b4a1
---

diff --git a/includes/parser/CoreParserFunctions.php b/includes/parser/CoreParserFunctions.php
index cf8cd41009..8a30ad1a22 100644
--- a/includes/parser/CoreParserFunctions.php
+++ b/includes/parser/CoreParserFunctions.php
@@ -178,7 +178,9 @@ class CoreParserFunctions {
 			default:
 				$func = 'urlencode';
 		}
-		return $parser->markerSkipCallback( $s, $func );
+		// See T105242, where the choice to kill markers and various
+		// other options were discussed.
+		return $func( $parser->killMarkers( $s ) );
 	}
 
 	public static function lcfirst( $parser, $s = '' ) {
diff --git a/tests/parser/parserTests.txt b/tests/parser/parserTests.txt
index 1670fe4cc5..c70f8b2ae6 100644
--- a/tests/parser/parserTests.txt
+++ b/tests/parser/parserTests.txt
@@ -19846,10 +19846,12 @@ Strip marker in urlencode
 {{urlencode:x<nowiki/>y}}
 {{urlencode:x<nowiki/>y|wiki}}
 {{urlencode:x<nowiki/>y|path}}
+{{urlencode:x<pre id="one">two</pre>y}}
 !! html
 <p>xy
 xy
 xy
+xy
 </p>
 !! end