From c398974918347e3491302d63177c787675f53e74 Mon Sep 17 00:00:00 2001
From: Michael Vogel <icarus@dabo.de>
Date: Mon, 8 Dec 2014 11:02:03 +0100
Subject: [PATCH 1/3] Logging to analyse the stale gprobe processes

---
 include/gprobe.php | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/include/gprobe.php b/include/gprobe.php
index 0cf32e95fe..36650eb9ae 100644
--- a/include/gprobe.php
+++ b/include/gprobe.php
@@ -10,7 +10,7 @@ function gprobe_run(&$argv, &$argc){
 	if(is_null($a)) {
 		$a = new App;
 	}
-  
+
 	if(is_null($db)) {
 	    @include(".htconfig.php");
     	require_once("include/dba.php");
@@ -37,6 +37,8 @@ function gprobe_run(&$argv, &$argc){
 		dbesc(normalise_link($url))
 	);
 
+	logger("gprobe start for ".normalise_link($url), LOGGER_DEBUG);
+
 	if(! count($r)) {
 
 		$arr = probe_url($url);
@@ -55,7 +57,8 @@ function gprobe_run(&$argv, &$argc){
 	}
 	if(count($r))
 		poco_load(0,0,$r[0]['id'], str_replace('/profile/','/poco/',$r[0]['url']));
-		
+
+	logger("gprobe end for ".normalise_link($url), LOGGER_DEBUG);
 	return;
 }
 

From 620ee6be0bdb4e706abfa2172d3dfb3e2d035e25 Mon Sep 17 00:00:00 2001
From: Michael Vogel <icarus@dabo.de>
Date: Mon, 8 Dec 2014 14:26:44 +0100
Subject: [PATCH 2/3] Cleanup the HTML before parsing it.

---
 include/Scrape.php | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/Scrape.php b/include/Scrape.php
index 99784af336..ad9e030a39 100644
--- a/include/Scrape.php
+++ b/include/Scrape.php
@@ -263,6 +263,11 @@ function scrape_feed($url) {
 	}
 
 	try {
+		// Cleanup invalid HTML
+		$doc = new DOMDocument();
+		@$doc->loadHTML($s);
+                $s = $doc->saveHTML();
+
 		$dom = HTML5_Parser::parse($s);
 	} catch (DOMException $e) {
 		logger('scrape_feed: parse error: ' . $e);

From 04106ff6f52b03d0d5b009f07a7b6ab11107f0c0 Mon Sep 17 00:00:00 2001
From: Michael Vogel <icarus@dabo.de>
Date: Mon, 8 Dec 2014 22:37:49 +0100
Subject: [PATCH 3/3] Moving the clean up code to a central place.

---
 include/Scrape.php       | 5 -----
 library/HTML5/Parser.php | 6 ++++++
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/include/Scrape.php b/include/Scrape.php
index ad9e030a39..99784af336 100644
--- a/include/Scrape.php
+++ b/include/Scrape.php
@@ -263,11 +263,6 @@ function scrape_feed($url) {
 	}
 
 	try {
-		// Cleanup invalid HTML
-		$doc = new DOMDocument();
-		@$doc->loadHTML($s);
-                $s = $doc->saveHTML();
-
 		$dom = HTML5_Parser::parse($s);
 	} catch (DOMException $e) {
 		logger('scrape_feed: parse error: ' . $e);
diff --git a/library/HTML5/Parser.php b/library/HTML5/Parser.php
index 5f9ca560e5..c7faf875ad 100644
--- a/library/HTML5/Parser.php
+++ b/library/HTML5/Parser.php
@@ -17,6 +17,12 @@ class HTML5_Parser
      * @return Parsed HTML as DOMDocument
      */
     static public function parse($text, $builder = null) {
+
+	// Cleanup invalid HTML
+	$doc = new DOMDocument();
+	@$doc->loadHTML($text);
+	$text = $doc->saveHTML();
+
         $tokenizer = new HTML5_Tokenizer($text, $builder);
         $tokenizer->parse();
         return $tokenizer->save();