[kudu] branch master updated: [util] set CURLOPT_NOPROXY for EasyCurl

[gr...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

granthenke pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/kudu.git


The following commit(s) were added to refs/heads/master by this push:
     new a33b095  [util] set CURLOPT_NOPROXY for EasyCurl
a33b095 is described below

commit a33b0951d91ecb0b84871bbf9d4b4b3550196af5
Author: Alexey Serbin <al...@apache.org>
AuthorDate: Mon Mar 23 23:41:05 2020 -0700

    [util] set CURLOPT_NOPROXY for EasyCurl
    
    As it turned out, some environments set 'http_proxy' and 'https_proxy'
    environment variables.  This makes results unreliable for test scenarios
    which fetch information from the embedded Kudu webserver.
    
    This patch addresses that by programmatically setting CURLOPT_NOPROXY
    to "*" wild-card by default, effectively disabling any proxying that
    might otherwise occur if the mentioned environment variables were set.
    
    Change-Id: If767bfb28516acb06f6243945300f5a362b869f8
    Reviewed-on: http://gerrit.cloudera.org:8080/15542
    Reviewed-by: Adar Dembo <ad...@cloudera.com>
    Tested-by: Kudu Jenkins
    Reviewed-by: Grant Henke <gr...@apache.org>
---
 src/kudu/util/curl_util.cc |  7 ++++++-
 src/kudu/util/curl_util.h  | 15 +++++++++++++++
 2 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/src/kudu/util/curl_util.cc b/src/kudu/util/curl_util.cc
index 82b0efc..e7992f7 100644
--- a/src/kudu/util/curl_util.cc
+++ b/src/kudu/util/curl_util.cc
@@ -71,7 +71,8 @@ size_t WriteCallback(void* buffer, size_t size, size_t nmemb, void* user_ptr) {
 #define CURL_RETURN_NOT_OK(expr) \
   RETURN_NOT_OK(TranslateError((expr), errbuf_))
 
-EasyCurl::EasyCurl() {
+EasyCurl::EasyCurl()
+    : noproxy_("*") {
   // Use our own SSL initialization, and disable curl's.
   // Both of these calls are idempotent.
   security::InitializeOpenSSL();
@@ -167,6 +168,10 @@ Status EasyCurl::DoRequest(const string& url,
         curl_, CURLOPT_CUSTOMREQUEST, custom_method_.c_str()));
   }
 
+  if (!noproxy_.empty()) {
+    CURL_RETURN_NOT_OK(curl_easy_setopt(curl_, CURLOPT_NOPROXY, noproxy_.c_str()));
+  }
+
   CURL_RETURN_NOT_OK(curl_easy_setopt(curl_, CURLOPT_HTTPAUTH, CURLAUTH_ANY));
   if (timeout_.Initialized()) {
     CURL_RETURN_NOT_OK(curl_easy_setopt(curl_, CURLOPT_NOSIGNAL, 1));
diff --git a/src/kudu/util/curl_util.h b/src/kudu/util/curl_util.h
index c1ddcc2..c05ce66 100644
--- a/src/kudu/util/curl_util.h
+++ b/src/kudu/util/curl_util.h
@@ -84,6 +84,19 @@ class EasyCurl {
     custom_method_ = std::move(m);
   }
 
+  // A comma-separated list of host names to avoid requests being proxied to,
+  // or "*" glob to disable proxying of any requests even if proxying is
+  // configured via CURLOPT_PROXY or '{http,https}_proxy' environment variables.
+  // An empty string "" clears the setting. By default, it's set to "*" since
+  // EasyCurl is primarily used in scenarios fetching data from embedded
+  // webservers of kudu-master/kudu-tserver running at the same host from where
+  // a request is issued, while 'http_proxy' and 'https_proxy' environment
+  // variables might be disruptive in that regard.
+  // See 'man CURLOPT_NOPROXY' for details.
+  void set_noproxy(std::string noproxy) {
+    noproxy_ = std::move(noproxy);
+  }
+
   // Whether to return an error if server responds with HTTP code >= 400.
   // By default, curl returns the returned content and the response code
   // since it's handy in case of auth-related HTTP response codes such as
@@ -111,6 +124,8 @@ class EasyCurl {
 
   std::string custom_method_;
 
+  std::string noproxy_;
+
   // Whether to verify the server certificate.
   bool verify_peer_ = true;