corrupted multibyte characters in node

[vkurchatkin]

Multibyte characters on chunk boundaries get corrupted.

[mzabriskie]

@vkurchatkin Would you mind providing example code so that I know how to reproduce this?

[vkurchatkin]

Here it is

var axios = require('axios');
var http = require('http');
var assert = require('assert');

var str = Array(100000).join('ж');

http.createServer(function(req, res) {
  res.setHeader('Content-Type', 'text/html; charset=UTF-8');
  res.end(str);
}).listen(4444, function() {
  axios.get('http://localhost:4444/').then(function(response) {
    process.nextTick(function() {
      assert(response.data === str);
    });
  });
});

I'm not sure that this test is 100% reliable (or that a 100% reliable test of this exists at all)

[mzabriskie]

Thanks. I'll give this a look.

[vkurchatkin]

Here is a quick fix:

diff --git a/lib/adapters/http.js b/lib/adapters/http.js
index 8a6fbf6..81da7ac 100644
--- a/lib/adapters/http.js
+++ b/lib/adapters/http.js
@@ -52,15 +52,15 @@ module.exports = function httpAdapter(resolve, reject, config) {
   // Create the request
   var transport = parsed.protocol === 'https:' ? https : http;
   var req = transport.request(options, function (res) {
-    var responseText = '';
+    var responseBuffer = [];
     res.on('data', function (chunk) {
-      responseText += chunk;
+      responseBuffer.push(chunk);
     });

     res.on('end', function () {
       var response = {
         data: transformData(
-          responseText,
+          Buffer.concat(responseBuffer).toString('utf8'),
           res.headers,
           config.transformResponse
         ),
@@ -83,4 +83,4 @@ module.exports = function httpAdapter(resolve, reject, config) {

   // Send the request
   req.end(data);
-};
\ No newline at end of file
+};

[mzabriskie]

@vkurchatkin thank you for your diff. I will apply this and write some tests around it.