fetch2/wget: fallback to GET if HEAD is rejected in checkstatus()

The core change here is to fall back to GET requests if HEAD is rejected in the
checkstatus() method, as you can't do a HEAD on Amazon S3 (used by Github
archives).  This meant removing the monkey patch that the default method was GET
and adding a fixed redirect handler that doesn't reset to GET.

Also, change the way the opener is constructed from an if/elif cluster to a
conditionally constructed list.

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 6ec70d5d2e330b41b932b0a655b838a5f37df01e)
Signed-off-by: Armin Kuster <akuster@mvista.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

2 files changed, 51 insertions, 23 deletions

diff --git a/lib/bb/fetch2/wget.py b/lib/bb/fetch2/wget.py
index bd2a8972a..0ff5b2ba3 100644
--- a/lib/bb/fetch2/wget.py
+++ b/lib/bb/fetch2/wget.py
@@ -234,38 +234,64 @@ class Wget(FetchMethod):
 
             return exported
 
-        def head_method(self):
-            return "HEAD"
-
+        class HTTPMethodFallback(urllib2.BaseHandler):
+            """
+            Fallback to GET if HEAD is not allowed (405 HTTP error)
+            """
+            def http_error_405(self, req, fp, code, msg, headers):
+                fp.read()
+                fp.close()
+
+                newheaders = dict((k,v) for k,v in req.headers.items()
+                                  if k.lower() not in ("content-length", "content-type"))
+                return self.parent.open(urllib2.Request(req.get_full_url(),
+                                                        headers=newheaders,
+                                                        origin_req_host=req.get_origin_req_host(),
+                                                        unverifiable=True))
+
+            """
+            Some servers (e.g. GitHub archives, hosted on Amazon S3) return 403
+            Forbidden when they actually mean 405 Method Not Allowed.
+            """
+            http_error_403 = http_error_405
+
+            """
+            Some servers (e.g. FusionForge) returns 406 Not Acceptable when they
+            actually mean 405 Method Not Allowed.
+            """
+            http_error_406 = http_error_405
+
+        class FixedHTTPRedirectHandler(urllib2.HTTPRedirectHandler):
+            """
+            urllib2.HTTPRedirectHandler resets the method to GET on redirect,
+            when we want to follow redirects using the original method.
+            """
+            def redirect_request(self, req, fp, code, msg, headers, newurl):
+                newreq = urllib2.HTTPRedirectHandler.redirect_request(self, req, fp, code, msg, headers, newurl)
+                newreq.get_method = lambda: req.get_method()
+                return newreq
         exported_proxies = export_proxies(d)
 
+        handlers = [FixedHTTPRedirectHandler, HTTPMethodFallback]
+        if export_proxies:
+            handlers.append(urllib2.ProxyHandler())
+        handlers.append(CacheHTTPHandler())
         # XXX: Since Python 2.7.9 ssl cert validation is enabled by default
         # see PEP-0476, this causes verification errors on some https servers
         # so disable by default.
         import ssl
-        ssl_context = None
         if hasattr(ssl, '_create_unverified_context'):
-            ssl_context = ssl._create_unverified_context()
-
-        if exported_proxies == True and ssl_context is not None:
-            opener = urllib2.build_opener(urllib2.ProxyHandler, CacheHTTPHandler,
-                    urllib2.HTTPSHandler(context=ssl_context))
-        elif exported_proxies == False and ssl_context is not None:
-            opener = urllib2.build_opener(CacheHTTPHandler,
-                    urllib2.HTTPSHandler(context=ssl_context))
-        elif exported_proxies == True and ssl_context is None:
-            opener = urllib2.build_opener(urllib2.ProxyHandler, CacheHTTPHandler)
-        else:
-            opener = urllib2.build_opener(CacheHTTPHandler)
-
-        urllib2.Request.get_method = head_method
-        urllib2.install_opener(opener)
-
-        uri = ud.url.split(";")[0]
+            handlers.append(urllib2.HTTPSHandler(context=ssl._create_unverified_context()))
+        opener = urllib2.build_opener(*handlers)
 
         try:
-            urllib2.urlopen(uri)
-        except:
+            uri = ud.url.split(";")[0]
+            r = urllib2.Request(uri)
+            r.get_method = lambda: "HEAD"
+            opener.open(r)
+        except urllib2.URLError as e:
+            # debug for now to avoid spamming the logs in e.g. remote sstate searches
+            logger.debug(2, "checkstatus() urlopen failed: %s" % e)
             return False
         return True
 
diff --git a/lib/bb/tests/fetch.py b/lib/bb/tests/fetch.py
index 7a08be0f4..3b74de0d7 100644
--- a/lib/bb/tests/fetch.py
+++ b/lib/bb/tests/fetch.py
@@ -737,6 +737,8 @@ class FetchCheckStatusTest(FetcherTest):
                       "ftp://ftp.gnu.org/gnu/autoconf/autoconf-2.60.tar.gz",
                       "ftp://ftp.gnu.org/gnu/chess/gnuchess-5.08.tar.gz",
                       "ftp://ftp.gnu.org/gnu/gmp/gmp-4.0.tar.gz",
+                      # GitHub releases are hosted on Amazon S3, which doesn't support HEAD
+                      "https://github.com/kergoth/tslib/releases/download/1.1/tslib-1.1.tar.xz"
                       ]
 
     if os.environ.get("BB_SKIP_NETTESTS") == "yes":