From 713b38e34c6d4a232af871659793bb9674e020f7 Mon Sep 17 00:00:00 2001
From: Frédéric Guillot <fred@miniflux.net>
Date: Fri, 19 Jan 2018 22:42:55 -0800
Subject: Handle more encoding edge cases

- Feeds with charset specified only in Content-Type header and not in XML document
- Feeds with charset specified in both places
- Feeds with charset specified only in XML document and not in HTTP header
---
 http/client.go   |  3 ++-
 http/response.go | 19 +++++++++++++++++--
 2 files changed, 19 insertions(+), 3 deletions(-)

(limited to 'http')

diff --git a/http/client.go b/http/client.go
index cdff0ce..b8541a2 100644
--- a/http/client.go
+++ b/http/client.go
@@ -95,11 +95,12 @@ func (c *Client) executeRequest(request *http.Request) (*Response, error) {
 		ContentLength: resp.ContentLength,
 	}
 
-	logger.Debug("[HttpClient:%s] OriginalURL=%s, StatusCode=%d, ContentLength=%d, ETag=%s, LastModified=%s, EffectiveURL=%s",
+	logger.Debug("[HttpClient:%s] OriginalURL=%s, StatusCode=%d, ContentLength=%d, ContentType=%s, ETag=%s, LastModified=%s, EffectiveURL=%s",
 		request.Method,
 		c.url,
 		response.StatusCode,
 		resp.ContentLength,
+		response.ContentType,
 		response.ETag,
 		response.LastModified,
 		response.EffectiveURL,
diff --git a/http/response.go b/http/response.go
index d9e9db6..a0cfc3f 100644
--- a/http/response.go
+++ b/http/response.go
@@ -6,8 +6,10 @@ package http
 
 import (
 	"io"
+	"mime"
 	"strings"
 
+	"github.com/miniflux/miniflux/logger"
 	"golang.org/x/net/html/charset"
 )
 
@@ -45,9 +47,22 @@ func (r *Response) IsModified(etag, lastModified string) bool {
 }
 
 // NormalizeBodyEncoding make sure the body is encoded in UTF-8.
+//
+// If a charset other than UTF-8 is detected, we convert the document to UTF-8.
+// This is used by the scraper and feed readers.
+//
+// Do not forget edge cases:
+// - Some non-utf8 feeds specify encoding only in Content-Type, not in XML document.
 func (r *Response) NormalizeBodyEncoding() (io.Reader, error) {
-	if strings.Contains(r.ContentType, "charset=") {
-		return charset.NewReader(r.Body, r.ContentType)
+	_, params, err := mime.ParseMediaType(r.ContentType)
+	if err == nil {
+		if enc, found := params["charset"]; found {
+			enc = strings.ToLower(enc)
+			if enc != "utf-8" && enc != "utf8" && enc != "" {
+				logger.Debug("[NormalizeBodyEncoding] Convert body to UTF-8 from %s", enc)
+				return charset.NewReader(r.Body, r.ContentType)
+			}
+		}
 	}
 	return r.Body, nil
 }
-- 
cgit v1.2.3