| 5dcc3ce8cae450f7800570b2826f5232096159cd |
|
20-Jan-2017 |
Yi Kong <yikong@google.com> |
DO NOT MERGE: Remove StuckServer using backlog and tests relying on it
This cherry-pick contains additional changes to remove more references
to StuckServer. The URLConnectionTest#teestConnectTimeouts() test is
suppressed in M via libcore/expectations/brokentests.txt and is not
expected to work. This change should not be merged into N branches.
(Original commit description follows below.)
StuckServer uses kernel implementation details that have changed on
version > 4.4, and causing tests depending on it to fail. Since we can't
find an alternative method to simulate a stuck server without special
permission, and the test itself is not reliable in that they don't fail
on non-stuck servers, remove them from Libcore tests.
Test: CtsLibcoreTestCases
Test: rm -rf out; m -j30 cts
Bug: 32551747
Bug: 35934294
(cherry picked from commit b5c7cd37726154f8da123e1844b5d60a63a60a2c)
Change-Id: I4dd951e9527b5ee5701c0c3ad3eaa0c6f34e8ef7
|
| 3ac6579c2666301209dcb116d60de4ba99d86e09 |
|
07-Feb-2017 |
Tobias Thierer <tobiast@google.com> |
Expand test coverage for disconnect during connect.
Before this CL, URLConnectionTest.testDisconnectDuringConnect()
tested the case of a HttpURLConnection being disconnected from
the CookieManager while connecting via getInputStream(). The
test missed the bug from http://r.android.com/329002 because
that bug only occurred during getOutputStream(), which was not
tested.
After this CL, connection initiation via the follow methods is
now covered:
- getInputStream()
- getOutputStream()
- getResponseCode()
- getResponseMessage()
Note that getErrorStream() is not used because it is documented
not to initiate a connection.
Bug: 34695782
Test: libcore.java.net.URLConnectionTest
Test: Checked that testDisconnectDuringConnect_getInputStream()
would have found the bug in http://r.android.com/329002
(unsubmitted) because it hangs indefinitely when run against
a build that includes that CL.
Change-Id: I6432a69b2cdac177f01e1ba84f32b99322e7c4e1
|
| f25faf811d480064e13b6bd3b8990d432d1a6510 |
|
24-Jan-2017 |
Tobias Thierer <tobiast@google.com> |
Test URLConnection.disconnect() while connecting
A HttpURLConnection.disconnect() while establishing a connection
could lead to an infinite busy loop. This CL adds a regression
test that exercises this behavior by disconnecting a connection
from inside the CookieHandler; the CookieHandler is consulted
before the actual connection is established, but late enough
to expose the original bug.
Bug: 33763156
Test: URLConnectionTest
Change-Id: I8e92f8830d6d62922a670e41d612d05f5f8d26b7
|
| e1d31ba3ac95b84d12ac24fae5561e3ac1dc66d7 |
|
18-Oct-2016 |
Tobias Thierer <tobiast@google.com> |
Update URLConnectionTest due to OkHttp bugfix backport
Applies changes to URLConnectionTest required for a corresponding
external/okhttp commit (backporting upstream commit
7ceaa2387f0be853222f5a1496f1f743fa6f8c6d) that fixes two bugs:
(1.) when attempting to tunnel a https connection over http, the host
header now include the :443, like it should. URLConnectionTest
was previously incorrectly asserting that the :443 was missing;
this CL fixes that assertion.
(2.) A bug in connecting to URLs with literal IPv6 addresses has been
fixed. This probably fixes bug 31917371 but this has not yet been
empirically verified. This CL adds a test to guard against
regression.
More details for (2.):
Between commit c358656c8799d30fd422448153e99a5dd37e298a (OkHttp 2.7.0)
and 7ceaa2387f0be853222f5a1496f1f743fa6f8c6d (OkHttp 3.2.0), upstream
OkHttp had a bug where a roundtrip HttpUrl -> host -> HttpUrl would
fail for literal IPv6 address hosts because the getter would return
the address without [] but the setter would only accept it wrapped in [].
Since Android's HttpURLConnection is derived from OkHttp 2.7.5, this
caused a crash when attempting a HttpURLConnection to a literal IPv6 URL.
Test: cts-tradefed run cts -m CtsLibcoreTestCases
Bug: 31917371
Change-Id: Ib4ee91356e7f11fe6f55e1cfd8f319f714a4567a
|
| 0b75bdc577edfee86b48265a945f383dffab100f |
|
18-Oct-2016 |
Tobias Thierer <tobiast@google.com> |
Test change for drop of insecure TLS protocol version fallback.
HttpsURLConnection no longer attempts insecure TLS protocol version
fallback; this CL changes the test's expectations accordingly.
Bug: 30185435
Bug: 30694151
Test: cts-tradefed run cts -m CtsLibcoreTestCases \
-t libcore.java.net.URLConnectionTest
Test: cts-tradefed run cts -m CtsLibcoreTestCases -a arm64-v8a
Change-Id: I633b6e5de23aec842429ff4310565a714023df5b
|
| 694b7112cb230a0db090d2b8d35166c8ffb19e93 |
|
18-Oct-2016 |
Tobias Thierer <tobiast@google.com> |
Deduplicate code in URLConnectionTest through a helper method.
This is in preparation for a follow-up CL to modify the shared
code.
Test: cts-tradefed run cts -m CtsLibcoreTestCases \
-t libcore.java.net.URLConnectionTest
Change-Id: I89cce7ed18491f343fa4ee9d0ad721d0e7520541
|
| fc575bca313624a7cba20171fbaa625bbc852cd4 |
|
17-Oct-2016 |
Tobias Thierer <tobiast@google.com> |
DO NOT MERGE Relax N CTS about whether \t is allowed in header values
Android N's Http(s)URLConnection initially forbade '\t' occurring in header
values when it should be allowed; this was fixed by [1] which will be
part of a future Android release. It's permitted for vendors shipping
Android N to apply this fix. Therefore, this CL relaxes CTS for Android N
to allow either behavior. CTS for APIs > 23 will enforce that \t is
allowed.
[1] AOSP commit dcabb633be0f71e720a856e141867d5d23adff93
Test: vogar libcore.java.net.URLConnectionTest#testRequestHeaderValidation
against (a) N (b) a recent build with http://r.android.com/272102
During local testing, temporarily added log statements after the two
asserts to verify that the correct branch was taken on each device.
Bug: 30799514
Change-Id: I4f37246337045574f5ab498d5e98b9c98e6feb74
|
| f14b48abd257bddff050624b51a73440689fc22f |
|
10-Aug-2016 |
Neil Fuller <nfuller@google.com> |
Modify tests to reflect different URL encoding for single quote
Single quotes in query parameters in HttpURLConnection changed
behavior between M and N due to upstream changes in OkHttp:
in N, single quotes started being encoded as %27 when present in
query parameters. Unfortunately, this broke several "iTunes remote"
apps: iTunes appears not to support %27 encoding for single quotes
in its DAAP server.
Bug: 30405333
Test: (old) cts test packages android.core.tests.libcore.package.okhttp
and android.core.tests.libcore.package.libcore
(cherry picked from commit 272317c3807fd40b04353354b540a6e2719e93a8)
Change-Id: I3d19dd880202bdc8e5909482e0e6c5167024d410
|
| 8f2073552a3aa97ce95f8380f5cab1742d77de91 |
|
14-Sep-2016 |
Chad Brubaker <cbrubaker@google.com> |
Drop SSLv3 support
The specific API version in the deprecation doc comments will be added
later, for now listed as TBD.
Bug: 30977793
Test: libcore/run-libcore-tests libcore/luni/src/test/java/libcore/javax/net/ssl/*
Change-Id: I5ebe22421589415a9e9065715262e6860b251b2b
|
| 3dd11b3751ef70fb35a629e244911b6f08a6afcb |
|
09-Sep-2016 |
Tobias Thierer <tobiast@google.com> |
Update test to allow HTAB in HTTP header values
Bug: 30799514
Test: run cts -m CtsLibcoreTestCases -t libcore.java.net.URLConnectionTest
Change-Id: Iba1cb2f52633d112e2d5a79fd11181c53913f6e2
(cherry picked from commit dcabb633be0f71e720a856e141867d5d23adff93)
|
| dcabb633be0f71e720a856e141867d5d23adff93 |
|
09-Sep-2016 |
Tobias Thierer <tobiast@google.com> |
Update test to allow HTAB in HTTP header values
Bug: 30799514
Test: run cts -m CtsLibcoreTestCases -t libcore.java.net.URLConnectionTest
Change-Id: Iba1cb2f52633d112e2d5a79fd11181c53913f6e2
|
| 272317c3807fd40b04353354b540a6e2719e93a8 |
|
10-Aug-2016 |
Neil Fuller <nfuller@google.com> |
Modify tests to reflect different URL encoding for single quote
Single quotes in query parameters in HttpURLConnection changed
behavior between M and N due to upstream changes in OkHttp:
in N, single quotes started being encoded as %27 when present in
query parameters. Unfortunately, this broke several "iTunes remote"
apps: iTunes appears not to support %27 encoding for single quotes
in its DAAP server.
Bug: 30405333
Test: (old) cts test packages android.core.tests.libcore.package.okhttp
and android.core.tests.libcore.package.libcore
Change-Id: Id87da28529098f639d3be6aed0626d5e3996739d
|
| 0f2730efd5fa2dbc1d7cdfb3d33b5b499eddb8c0 |
|
29-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Proxy connect: Test expected normalization for URLs with empty path
This is to prevent accidental future changes to this behavior detail.
Test: cts-tradefed run cts -c libcore.java.net.URLConnectionTest
Bug: 30107354
Change-Id: Idd7414c75adb9f76c85a6f5cc31c783faf99dfd9
|
| a9aa7271b66b750a23d102ad82bec50a131889c5 |
|
29-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Revert "Regression test for HTTP proxy request line"
This reverts commit c09900cee5e9f518ff8faa954044ba063a5a3e4e.
Change-Id: I2b93e200cb709e2ee2ed18dc90b898a273f8125b
|
| c09900cee5e9f518ff8faa954044ba063a5a3e4e |
|
13-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Regression test for HTTP proxy request line
Adds a test that demonstrates a regression from M in
the HTTP request line when a proxied connection
is made. It only affects URLs with empty paths (i.e. ones
without even a '/' after the host / authority). For
example, on M the request line for a request to
new URL(http://myhost).openConnection() would be:
GET http://myhost HTTP/1.1
but on N, without this change, it would be:
GET http://myhost/ HTTP/1.1
This change reverts to the M behavior.
Bug: 30107354
Bug: 29983827
Change-Id: Iea34afe3beb67022f746aa1b04262e43d841aacd
(cherry picked from commit d1288ce66ea8cb60353e8b1db9d31c8d5cf28127)
|
| d2e78a951c29949bc1cbb691c42ea4202340f138 |
|
12-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Regression test for HTTP proxy request line
Adds a test that demonstrates a regression from M in
the HTTP request line when a proxied connection
is made. It only affects URLs with empty paths (i.e. ones
without even a '/' after the host / authority). For
example, on M the request line for a request to
new URL("http://myhost").openConnection() would be:
GET http://myhost HTTP/1.1
but on N, without this change, it would be:
GET http://myhost/ HTTP/1.1
This change reverts to the M behavior.
Bug: 29983827
Change-Id: Ic457230cb9808546e079680b7ace79423ec30544
(cherry picked from commit d1288ce66ea8cb60353e8b1db9d31c8d5cf28127)
|
| d1288ce66ea8cb60353e8b1db9d31c8d5cf28127 |
|
12-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Regression test for HTTP proxy request line
Adds a test that demonstrates a regression from M in
the HTTP request line when a proxied connection
is made. It only affects URLs with empty paths (i.e. ones
without even a '/' after the host / authority). For
example, on M the request line for a request to
new URL("http://myhost").openConnection() would be:
GET http://myhost HTTP/1.1
but on N, without this change, it would be:
GET http://myhost/ HTTP/1.1
This change reverts to the M behavior.
Bug: 29983827
Change-Id: Ic457230cb9808546e079680b7ace79423ec30544
|
| 4869d3652a33a16bf96ea4685845c5efd337a7e7 |
|
04-Jul-2016 |
Tobias Thierer <tobiast@google.com> |
Fix URLConnectionTest asserting on the wrong connection
This was introduced in 321eed87d5dc289ea7c212c2fb743f57cc8d5b15.
I've moved the declarations such that other variable for the
other connection is not visible in each block.
Test: run cts -c libcore.java.net.URLConnectionTest
Change-Id: Idc028b7ee19b93b1431518fde1fd5e290ca8a128
|
| 321eed87d5dc289ea7c212c2fb743f57cc8d5b15 |
|
02-Jun-2016 |
Tobias Thierer <tobiast@google.com> |
URLConnectionTest changes for upgrade to OkHttp 2.7.5
These changes are required due to behavioral changes between OkHttp
2.5 and 2.7.5 which we want to accept for AOSP.
- ProxySelectors now get a URL reconstructed from an Address (without
path, query parameters and fragment) rather than the full request
URL. Proxy selection should not usually depend on those values:
DefaultProxySelector cannot be configured to do so;
PacProxySelector probably can be but it'd be a hack. In the absence
of evidence that a significant number of clients rely on such a
hack, we don't want to try to continue to support such behavior.
This behavioral change affects testRedirectWithInvalidUrl() and was
introduced by:
github.com/square/okhttp/commit/c358656c8799d30fd422448153e99a5dd37e298a
which no converts the URL to an Address, then back to a URL that's
delivered to the proxy selector.
- When clients set a Proxy-Authorization header on a request to a
target server (e.g. google.com), previous versions of OkHttp would
have copied that header to the request to the proxy that
established a tunnel. It's unreasonable to support clients that may
rely on this hack because this leaks the Proxy-Authorization header
to the target server. Instead, the Proxy-Authorization header is no
generated on the library level in response to a HTTP 407 Proxy
authentication required status, which requires an additional round
trip.
Similarly, the request to the proxy to set up the tunnel used to
have the User-Agent copied from the initial tunneled request.
Instead, this request's User-Agent is now generated at the library
level through Version.userAgent()
- OkHttp 2.7 has new logic for CertificatePinning that we do not
want to expose through Android's API surface; Android already
has a different implementation of certificate pinning. This CL
adds a sanity check that some Platform logic that OkHttp 2.7.5
uses only for certificate pinning is not invoked in the common
case of Http{s}URLConnections.
More details on these and other changes that didn't affect tests, at:
https://docs.google.com/document/d/19PF3Exd_q32gAGCiRFWRf0Pq_xrIWs-cRViHkFTxJg8/edit
Change-Id: I29277aea0beb8921b3038fefb922b669e01bf106
|
| c3ccff57d6029c0105eb7b100708d0b4883c6160 |
|
15-Jun-2016 |
Tobias Thierer <tobiast@google.com> |
Fix URLConnectionTest.testConnectTimeouts() by dropping use of Mockito
This test used a Mockito spy to wrap a Socket instance.
It has been failing on AOSP for (at least) weeks due to what
appears like a bug in AOSP's version of Mockito; this bug can lead
to crashes in <clinit>, or the expectations set up on the spy not
working correctly. There is probably some way we could call the
Mockito API to cause the static initializers to run in a different
order and work around the bug, but that would require more
investigation and would likely still be fragile. It doesn't seem
like a good trade-off as long as this only affects isolated cases.
Therefore this CL replaces the spy with an explicit DelegatingSocket
wrapper instead. This avoids the need to investigate a potentially
fragile workaround for the Mockito behavior and increases clarity by
removing the runtime magic.
Bug: 29095428
Change-Id: I46d57f18d9ea157e8a34ce3aead55e0f97de458a
|
| cc4ece7f7ff42e955ac454aed51fc1a8e38abe4c |
|
15-Jun-2016 |
Tobias Thierer <tobiast@google.com> |
URLConnectionTest: explicitly close() all the TestSSLContexts
This fixes the test failing from the un-close()d TestSSLContexts
when run on vogar. It doesn't disconnect Http{,s}URLConnections
which the test constructs; these can cause the test to fail
similarly.
The un-close()d contexts caused the test to fail in this manner:
- TestSSLContext.serverSocket.getImpl() instanceof SocksSocketImpl
which extends AbstractPlainSocketImpl.
- That base class in turn uses CloseGuard to enforce (best effort)
that close() is being called.
I originally observed this in vogar on a local branch with an
updated version of OkHttp (2.7.5). I have reproduced it on
AOSP master head but have not reproduced it under cts-tradefed.
Therefore I conclude that there is an additional bug that caused
this test to pass in cts-tradefed when it should have failed; I
have filed bug 29533469 for that.
Change-Id: Ifffeb576b25e07375917dcfc0fdec3c6a9da1be9
|
| 5ccc6fa33dd097cb501869d041bef20e45c8d47d |
|
22-Jun-2016 |
Tobias Thierer <tobiast@google.com> |
Disable leak detection in URLConnectionTest by extending TestCase.
URLConnectionTest previously extended
AbstractResourceLeakageDetectorTestCase which supplied the
leak detection logic, but that logic was unreliable.
This created more boilerplate and maintenance work in unrelated
tests (which are not focused on leak detection) exceeding the
value it provides:
- The leak detection doesn't work at all under CTS (29533469)
and is very flaky on vogar.
- Currently URLConnectionTest constructs 19 instances of
TestSSLContext and 127 instances of Http(s)URLConnection,
most of which aren't closed/disconnected. Closing all of
these would add excessive boilerplate and pollute the
purpose of these tests.
Tested:
- Test now passes under vogar
Bug: 29533469
Change-Id: I5ef6a51e651b52d975a4cd0dfe4548b83a328c72
|
| 41a4ba666dcb6eb5ad498fe133c8022afdf0c935 |
|
20-Jun-2016 |
Tobias Thierer <tobiast@google.com> |
Revert "URLConnection: Add regression test for b/26769689."
WhatsApp have address the issue so we no longer need this
workaround.
This reverts commit c7b148fd04d1f7de9ced0d56a97014467f86a59b.
Bug: 27353040
Change-Id: Ie4aa1187cc602eb9d2123d58deca9034ee6f2021
|
| 9f6ca97141aa1094a447f78a196988a2d52c438f |
|
25-May-2016 |
Neil Fuller <nfuller@google.com> |
Relax tests around HTTP header validation
An associated change in OkHttp is relaxing the validation to
make it closer to the behavior in M. This may be temporary and
could be removed in a future release of Android.
Bug: 28867041
Bug: https://code.google.com/p/android/issues/detail?id=210205
(cherry picked from commit e62b4c5890b48404714004e3d8a167479222b87b)
Change-Id: I8e2e58a583979611d78f9de6e24bd3c5a7dc2879
|
| e62b4c5890b48404714004e3d8a167479222b87b |
|
25-May-2016 |
Neil Fuller <nfuller@google.com> |
Relax tests around HTTP header validation
An associated change in OkHttp is relaxing the validation to
make it closer to the behavior in M. This may be temporary and
could be removed in a future release of Android.
Bug: 28867041
Bug: https://code.google.com/p/android/issues/detail?id=210205
Change-Id: If45f89578553e38ad455850f7f6bbecd4d51262e
|
| 7abb54e68e213538f8df3357a8abce4a6e49b086 |
|
05-Apr-2016 |
Sergio Giro <sgiro@google.com> |
TestKeyStore: add the ability to set certificate serial numbers
Change URLConnectionTest#testHttpsWithCustomTrustManager
to check for specifically set serial numbers.
As of 5a85130cc3ee3df65c3b263773e3649277b37317, the serial numbers
were being generated randomly, and the test was looking for the
value "1", so the test was broken.
(cherry picked from commit a23e1a7d40a0a85f8eed3a94d6b5714f98180a4e)
Bug: 27987415
Change-Id: I91d2726f362a6aab75edbb6b0d133c00ab774e73
|
| a23e1a7d40a0a85f8eed3a94d6b5714f98180a4e |
|
05-Apr-2016 |
Sergio Giro <sgiro@google.com> |
TestKeyStore: add the ability to set certificate serial numbers
Change URLConnectionTest#testHttpsWithCustomTrustManager
to check for specifically set serial numbers.
As of 5a85130cc3ee3df65c3b263773e3649277b37317, the serial numbers
were being generated randomly, and the test was looking for the
value "1", so the test was broken.
Bug: 27987415
Change-Id: I1b6a8dbcb496808266f66ede9503027ab8693861
|
| 72d80eec6c1516cf5d9c2b7746b8e32f43eb5107 |
|
16-Mar-2016 |
Neil Fuller <nfuller@google.com> |
Fix for a runtime exception that could occur with bad redirects
If the redirect address contained an invalid % escape sequence
the request would fail with a runtime exception when the URI
object was created to be passed to ProxySelector. Now the
bad escape sequence is itself escaped. e.g. %0& -> %250&. The
actual request made is unchanged (i.e. it will still contain
%0&).
This is related to changes in external/okhttp.
Bug: 27590872
(cherry picked from commit 7ac2143d73aa65c1d8bdb448ad11ee90a5503a85)
Change-Id: Ic2a5173edb817822c185694d7b64e9e3a2a56b91
|
| 7ac2143d73aa65c1d8bdb448ad11ee90a5503a85 |
|
16-Mar-2016 |
Neil Fuller <nfuller@google.com> |
Fix for a runtime exception that could occur with bad redirects
If the redirect address contained an invalid % escape sequence
the request would fail with a runtime exception when the URI
object was created to be passed to ProxySelector. Now the
bad escape sequence is itself escaped. e.g. %0& -> %250&. The
actual request made is unchanged (i.e. it will still contain
%0&).
This is related to changes in external/okhttp.
Bug: 27590872
Change-Id: I17e00aed33ff77c2f527e78f311c47c7a1c6bdb6
|
| 2325d15f20e54996573fe6656320db317fd28103 |
|
11-Mar-2016 |
Sergio Giro <sgiro@google.com> |
URLConnectionTest: add workaround for \r
Bug: 26422335
Bug: 26889631
Bug: 27606665
(cherry picked from commit 0f1529a265de32624ea0a675bd54b7861859c496)
Change-Id: If90a1f828abf251ad6a8a37c91d91cf17e6d3844
|
| 0f1529a265de32624ea0a675bd54b7861859c496 |
|
11-Mar-2016 |
Sergio Giro <sgiro@google.com> |
URLConnectionTest: add workaround for \r
Bug: 26422335
Bug: 26889631
Bug: 27606665
Change-Id: I88f5c0d456b871eccb36b49eb22e09065258ec02
|
| c7b148fd04d1f7de9ced0d56a97014467f86a59b |
|
15-Feb-2016 |
Narayan Kamath <narayan@google.com> |
URLConnection: Add regression test for b/26769689.
IPv6 Literal addresses must appear with surrounding square-brackets.
bug: 26769689
Change-Id: I2bf4c7b73adb8bc2a18a185771fa96742e8579d5
|
| 80d240bbef2cab434191cd4e60f45d117dcf8c92 |
|
26-Jan-2016 |
Narayan Kamath <narayan@google.com> |
URLConnectionTest: Change character mapping expectations.
Track commit 65f07bec4f480898b1c41472234b833497482829
in external/okhttp.
bug: 26455836
bug: 26843611
Change-Id: I71da6d9ebd07aecda7b8ca00e3a850c3ca2e2eca
|
| 9a649c0fa7636ffa21786a9c78e43effd3b0f457 |
|
12-Jan-2016 |
Neil Fuller <nfuller@google.com> |
Tests for request header validation
Includes tests for the workaround added for bug so it is not
lost prematurely.
Bug: 26422335
Change-Id: Ic4bf7511bfa2e323ee05522b5da0a9141364f2e3
|
| 8821bc671060ee45e95fb6da6694645c15d7c577 |
|
30-Nov-2015 |
Neil Fuller <nfuller@google.com> |
Add tests for "follow redirects" behavior
Bug: https://code.google.com/p/android/issues/detail?id=194495
Bug: 25928343
Change-Id: Ic0ff2f94f7de0caa0e5c76be7443b0742cb96977
|
| f6e698d17090d55159e19dd15a74df953f8ade63 |
|
01-Sep-2015 |
Neil Fuller <nfuller@google.com> |
Changes associated with upstream OkHttp pull
Contains changes related to how bad characters,
particularly in hostnames, are handled. These
are all related to OkHttp's changes to use HttpUrl
internally. URL and URI were previously used and
were less fussy.
The following tests changed:
URLConnectionTest.testHostWithNul():
Hosts with nul (\0) characters in them
throw UnknownHostException instead of
IllegalArgumentException. This is consistent
with the behavior with other bad host names
(like those with spaces which have historically
been treated as UnknownHostException).
A checked exception generally seems better than
an unchecked one.
URLConnectionTest.testUrlWithSpaceInHostViaHttpProxy():
Hosts with spaces in them returned by proxies will
now throw an UnknownHostException, not nothing.
This reverts the behavior to as it was in older
versions of Android (see commit
ac03ce489cbe22905ee525e005d67df0a74ff8de).
URLConnectionTest.testUrlCharacterMapping():
Hosts are now canonicalized to lower-case.
Hosts can now be rejected if the characters are
invalid and outside of supported IDN input codes.
Some characters in paths and queries are no longer
supported, will not be encoded or may be encoded
differently. e.g. '@' is no longer encoded in paths,
'[' is no longer supported.
Bug: 23553256
Bug: 22943975
Change-Id: Icfbcd520fd4e861dc6a0261aba982ee44d9f80cc
|
| a77f9f561d6bf7c2fb789b8065f7017351f1cacd |
|
14-Sep-2015 |
Neil Fuller <nfuller@google.com> |
Add additional tests around HttpURLConnection character handling
The upcoming OkHttp upgrade modifies the behavior. It will be
easier to debate / document what is changing by adding tests that
cover the existing behavior.
Bug: 23553256
Change-Id: Ia7e55bfd2685a6914269cfcb7deb95c78fc20db0
|
| 26f2557b26ea23326178f029e07a8adbfc27d0bf |
|
08-Sep-2015 |
Neil Fuller <nfuller@google.com> |
Fix / unsuppress some URLConnectTests tests
URLConnectionTest.testConnectTimeouts():
The code was creating a URL like:
http://:::<port number>/
because the stuck server returns "::" as the host
address for the wildcard binding done. This URL
is invalid (as it is missing [] around the IPv6
address string).
The test was supposed to check multiple IPs are tried
and the overall timeout is the sum of all the sockets
tried.
Switching StuckServer over to binding to localhost
did not work.
Since the format of the URL does not appear to be the
purpose of this test it is reasonable to change it
before the OkHttp update.
Instead, the test now replaces the the SocketFactory
with one that returns sockets that immediately time
out, avoid the need to nerf the server-side entirely.
The test does not test multiple IPs because (AFAIK) the
tests cannot depend on being able to resolve real
server names. This part of the change involved creating
a hidden method in SocketFactory.
This change adds a new dependency on Mockito so that a
partial mock for Socket can be created.
URLConnectionTest.testServerShutdownInput():
Unsuppress. It should work now.
Bug: 5534202
Bug: 23888521
Change-Id: Iaa57987c59cd6ec1c117f9c90f6ff8ea42d58210
|
| 381dc32dba0a02497a29f267f11d074d4fe39bcd |
|
05-Aug-2015 |
Neil Fuller <nfuller@google.com> |
Revert "Revert "Changes associated with an OkHttp bad proxy response fix""
This reverts commit 11c3934f61c6c873fd9caae16947285bc0068965.
commit 11c3934f61c6c873fd9caae16947285bc0068965 was incorrectly
merged from lollipop-cts-dev into lollipop-mr1-cts-dev.
Change-Id: I66fa65f08397fb34d6af1cb73d2c0fa7f57ef1b6
|
| 11c3934f61c6c873fd9caae16947285bc0068965 |
|
17-Jul-2015 |
Neil Fuller <nfuller@google.com> |
Revert "Changes associated with an OkHttp bad proxy response fix"
This reverts commit 467a6809cd31cbfa94c326415ff3c441d9b1e8be.
commit 467a6809cd31cbfa94c326415ff3c441d9b1e8be is present in
lollipop-cts-dev but not lollipop-release. This causes the
tests to fail when run against a lollipop-release build.
DO NOT MERGE ANYWHERE
Bug: 22546532
Change-Id: I4582824e94679213f27352e18e7d9bb985954909
|
| e8e4c88dbbc82ca581a899a152a4ff7a13217232 |
|
26-Jan-2015 |
Neil Fuller <nfuller@google.com> |
Regression tests for Etag handling
Confirm that Etag headers are implemented correctly.
Bug: https://code.google.com/p/android/issues/detail?id=108949
Change-Id: Ida716d089ab1367ae955bb110a450c63ce3a1fa9
|
| d0ebcbd92bd1910c043a5b9cfc0c64f547f17ba6 |
|
23-Jan-2015 |
Neil Fuller <nfuller@google.com> |
Fix incorrect ordering of assertEquals
Change-Id: I69ddea2941400881ef64309d2470953d25136c1a
|
| 764ea3d867b544de457aac2d02327d7b796d4a5c |
|
02-Dec-2014 |
Neil Fuller <nfuller@google.com> |
Changes needed to support OkHttp upgrade
Change-Id: Ifd6ed8ab22e7fd5d8cb8c9eb28976291d125c0e4
|
| 467a6809cd31cbfa94c326415ff3c441d9b1e8be |
|
05-Nov-2014 |
Neil Fuller <nfuller@google.com> |
Changes associated with an OkHttp bad proxy response fix
See the associated external/okhttp that changes the behavior
when encountering a body response from a CONNECT request.
Bug: 6754912
Bug: 18575971
(cherry picked from commit 305af3892e7b62c80ce5888a2514141ea3008ae2)
Change-Id: Idaf1fb5ef7cd3395e1fe26280799321715509e94
|
| 4284daf6172310292947bd62540e3007164fdde9 |
|
27-Nov-2014 |
Narayan Kamath <narayan@google.com> |
Add a simple test for "last-modified" headers.
Tests that getLastModified() == getHeaderFieldDate("Last-Modified").
This was lost in commit d8ce757d85ba665fa61d35e25e03c4de16c0f4c6.
Change-Id: I0dd1e7091b690d70161f33b8dfe3d0cee7ee4c9b
|
| d8ce757d85ba665fa61d35e25e03c4de16c0f4c6 |
|
22-Nov-2014 |
Aaron Whyte <awhyte@google.com> |
Fix URLConnectionTest for devices without full internet access.
This removes two bad tests that are covered elsewhere, using
mocks.
test_getContentEncoding() was hitting http://www.amazon.com, to see
if a URLConnection getter was working. That getter is now tested in
libcore/luni/src/test/java/libcore/java/net/URLConnectionTest.java,
in testClientConfiguredGzipContentEncoding().
test_getLastModified() was hitting
http://www.php.net/manual/en/function.explode.php, to see if
getHeaderFieldDate was working for Last-Modified. That getter is
better tested in test_getHeaderFieldDateLjava_lang_StringJ(), in the
same file.
Bug:18480776
(cherry picked from commit 939e6a1aceff448445928ad5d815479e84e84383)
Change-Id: Ia01fd59809683317271017364179e1843c149e65
|
| 939e6a1aceff448445928ad5d815479e84e84383 |
|
22-Nov-2014 |
Aaron Whyte <awhyte@google.com> |
Fix URLConnectionTest for devices without full internet access.
This removes two bad tests that are covered elsewhere, using
mocks.
test_getContentEncoding() was hitting http://www.amazon.com, to see
if a URLConnection getter was working. That getter is now tested in
libcore/luni/src/test/java/libcore/java/net/URLConnectionTest.java,
in testClientConfiguredGzipContentEncoding().
test_getLastModified() was hitting
http://www.php.net/manual/en/function.explode.php, to see if
getHeaderFieldDate was working for Last-Modified. That getter is
better tested in test_getHeaderFieldDateLjava_lang_StringJ(), in the
same file.
Bug:18480776
Change-Id: I21ffc5c9c72eb4379478c11687d197dab176a851
|
| 5df4f660e3a115d450633c6c1e3a5220d3a9708d |
|
12-Nov-2014 |
Neil Fuller <nfuller@google.com> |
Test changes for the HttpURLConnection fallback strategy
The fallback strategy has changed to perform a multi-level
fallback, rather than trying
"default enabled protocols" and "SSLv3, if enabled".
Fallback now includes attempts leading with TLSv1.2, TLSv1.1,
TLSv1 and SSLv3 if they are enabled.
Bug: 13228108
Change-Id: I6361a4103461bffd50bf7849fe92cb7e384a1930
|
| ce0ccd0299a8c23666a630afa4e440963e457615 |
|
12-Nov-2014 |
Neil Fuller <nfuller@google.com> |
Fixes associated with OkHttp fallback changes
OkHttp now selects from the available, enabled protocols rather
than forcing SSLv3 to be on. Without this change, if SSLv3 is not
part of the protocols enabled by default on the device this test
would fail.
Bug: 13228108
Change-Id: I5b89a73af9cf729dbf892f958fa318daa4109ff4
|
| 1d99fe5087ee6a512bb2d38ce5b0f6e3bbc2ea79 |
|
12-Nov-2014 |
Neil Fuller <nfuller@google.com> |
Improve tests for TLS fallback.
This is a small improvement over the changes that were submitted to
Android when TLS_FALLBACK_SCSV was introduced.
Previously the test would cripple the server to use a low quality
protocol so the TLS_FALLBACK_SCSV check would pass for all connection
attempts.
This change modifies that approach. Instead it introduces a test
client socket factory that ignores any attempt to set the
TLS_FALLBACK_SCSV cipher. This means that the server will
not attempt to enforce the fallback check.
The ideal solution would be to disable TLS_FALLBACK_SCSV checks in
the MockWebServer for the tests that are supposed to simulate servers
that do not perform TLS_FALLBACK_SCSV checks. However, we do not
currently have a mechanism to do this.
This change also adjusts tests that test fallback to explicitly
enable SSLv3 for server-side socket factories. The Android default
was changed to disable SSLv3 (though it is still supported). An
up-coming change should remove the need for this.
Change-Id: I5bd6c2659ebbb8eff729f75b5208ca414d70e385
|
| 305af3892e7b62c80ce5888a2514141ea3008ae2 |
|
05-Nov-2014 |
Neil Fuller <nfuller@google.com> |
Changes associated with an OkHttp bad proxy response fix
See the associated external/okhttp that changes the behavior
when encountering a body response from a CONNECT request.
Bug: 6754912
Change-Id: I20553dda374cd4903dc83cceb8d203662319571e
|
| 25f20231a8e43b92d7128aa57e52a1d4d8e90fea |
|
24-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Update tests for HttpURLConnection.disconnect() changes
These changes are associated with external/mockwebserver
and external/okhttp commits.
URLConnectionTest#testDisconnectedConnection
- Affected by the bug fix.
URLConnectionTest#testSslFallback
- Affected by changes to MockWebServer. Handshake failures
are no longer queued/recorded.
HttpsURLConnectionTest
- API changes related to MockWebServer.
Bug: 18083851
(cherry pick from commit 246e740c85527012891ac3b07cb658dd2fcabc87)
Change-Id: Ia7fe3961b1bc28ef2efa69bbf59a16bc54d07063
|
| 246e740c85527012891ac3b07cb658dd2fcabc87 |
|
24-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Update tests for HttpURLConnection.disconnect() changes
These changes are associated with external/mockwebserver
and external/okhttp commits.
URLConnectionTest#testDisconnectedConnection
- Affected by the bug fix.
URLConnectionTest#testSslFallback
- Affected by changes to MockWebServer. Handshake failures
are no longer queued/recorded.
HttpsURLConnectionTest
- API changes related to MockWebServer.
Bug: 18083851
Change-Id: Iec211a1e1e713623ccc8f8ca1de79833abe2f961
|
| 7618d0b86f53ada960898d531e53e1980914539f |
|
03-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Changes to support TLS fallback SCSV
Fixes for tests.
(cherry picked from commit 957ec8b09833e1c2f2c21380e53c13c9962e6b3e)
Change-Id: I848b6c84524aabffb1dc71c00aae8f3e9b741ef0
|
| 6c9609af5f63a759bd50b5f6586f6f52502b4f93 |
|
07-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Add support for TLS_FALLBACK_SCSV
Backport of commits:
external/conscrypt: 8d7e23e117da591a8d48e6bcda9ed6f58ff1a375
libcore: e6a6e935e98f426c7000b2bf4086f87101f4441c
libcore: 957ec8b09833e1c2f2c21380e53c13c9962e6b3e
Plus additional changes to:
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLServerSocketImpl.java
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLSocketImpl.java
luni/src/test/java/libcore/java/net/URLConnectionTest.java
luni/src/test/java/org/apache/harmony/xnet/provider/jsse/CipherSuiteTest.java
support/src/test/java/libcore/java/security/StandardNames.java
to account for KitKat differences.
Bug: 17750026
Change-Id: Ic6e9474275bc3ffec3b5c2d6df1f8d6ffe77bff8
|
| 865c83f8383f03d545217c35d9140a4627a74406 |
|
07-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Add support for TLS_FALLBACK_SCSV
Backport of commits:
external/conscrypt: 8d7e23e117da591a8d48e6bcda9ed6f58ff1a375
libcore: e6a6e935e98f426c7000b2bf4086f87101f4441c
libcore: 957ec8b09833e1c2f2c21380e53c13c9962e6b3e
Plus additional changes to:
luni/src/main/java/libcore/net/http/HttpConnection.java
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLServerSocketImpl.java
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLSocketImpl.java
luni/src/test/java/libcore/java/net/URLConnectionTest.java
luni/src/test/java/org/apache/harmony/xnet/provider/jsse/CipherSuiteTest.java
support/src/test/java/libcore/java/security/StandardNames.java
to account for JellyBean differences.
Bug: 17750026
Change-Id: I7b60b1260fa0b275631ce9987168c7b5fc7ca138
|
| 957ec8b09833e1c2f2c21380e53c13c9962e6b3e |
|
03-Oct-2014 |
Neil Fuller <nfuller@google.com> |
Changes to support TLS fallback SCSV
Fixes for tests.
Change-Id: Ie8e01e8fa654afd6b6c9f30ec46552d46326e467
|
| fa023837b8be8e9eda6008398648b4c9b1aa1e3e |
|
28-May-2014 |
Neil Fuller <nfuller@google.com> |
Fix for flakiness in URLConnectionTest.testServerShutdownOutput
Flakiness reported by Junjie.Hu <junjie.hu@mediatek.com>
This fix relies on changes to MockWebServer to provide
deterministic ordering of responses.
Change-Id: I4f296fa2f8d227876f52c69e22001f4e0d7b4ffd
|
| a6f350c645dbb66d68cc2b03afb8f2eeaa88fbba |
|
28-Feb-2014 |
Paul Duffin <paulduffin@google.com> |
Improve detection of CloseGuard protected resource leakage
* Add CloseGuardMonitor to intercept and collate CloseGuard reports and if
necessary throw an exception listing the resource leaks.
* Add ResourceLeakageDetector to abstract away the CloseGuardMonitor which will
not work on RI.
* Add AbstractResourceLeakageDetectorTestCase as a base class for tests
that need to detect resource leaks, in future this could be handled by
modifications to Cts and Vogar test runners.
* Remove CloseGuardTester and its sole usage in ProcessBuilderTest.
* Remove CloseGuardGuard from within URLConnectionTest
* Change ZipFileTest, ProcessBuilderTest, URLConnectionTest to use new
mechanism, fix issues that are identified and do some cleanup/remove
duplicated code.
Bug: https://code.google.com/p/android/issues/detail?id=66383
Change-Id: Id026dbb6bc66091a15f07329e6371cd0d1f32cf5
|
| ac03ce489cbe22905ee525e005d67df0a74ff8de |
|
06-Mar-2014 |
Neil Fuller <nfuller@google.com> |
Relax testUrlWithSpaceInHostViaHttpProxy test
This is for the new OkHttp version. The older
version detected the URL is invalid on Android. The newer
version does not. The correct behavior is undefined and
this test was added to prove that NullPointerException is
not thrown, which it will still test for.
Change-Id: Ibbaaf006982eba9de2d3751c6ee302dd28f7786f
|
| 9749b98a92fbf2bfe095b7ec884789f62ab293d8 |
|
03-Mar-2014 |
Neil Fuller <nfuller@google.com> |
Fix URLConnectionTest#testSetChunkedStreamingMode
Commit 60f79119fd7e37c459a96888594f90d86bab5535 broke it
- until OkHttp is upgraded the chunkLength is obeyed.
The argument needs to be >= 8 for the test to pass.
Change-Id: I22acaa4a56fdbd22abc075a2300cf53e303ec791
|
| 60f79119fd7e37c459a96888594f90d86bab5535 |
|
27-Feb-2014 |
Neil Fuller <nfuller@google.com> |
Tests and docs updates for OkHttp update.
OkHttp has made various changes. This fixes the known
breakages before the update is applied.
1) Chunk length is no longer honored exactly. The
API docs have been updated to reflect this.
2) testConnectViaHttpsWithSSLFallback simulated an SSL
handshake failure by closing the socket. This is no longer
sufficient to trigger the fallback behavior after OkHttp change
46a0852. testSslFallback tests an actual handshake failure and
provides coverage. testConnectViaHttpsWithSSLFallback
has therefore been removed.
3) testRetryableRequestBodyAfterBrokenConnection is similar to
testConnectViaHttpsWithSSLFallback. Previously it passed because
of the SSL fallback behavior. That fallback no longer happens.
It is possible this was actually supposed to be checking the
internal RetryableOutputStream. Now OkHttp does not retry
connections after the connection is made it is not possible
to test the request body is "replayed" to the server.
4) testNonRetryableRequestBodyAfterBrokenConnection: Related to
3 above. It was testing that fixed length request bodies are not
retryable. This test continues to pass with OkHttp but only
because it fails to retry generally, not because of the request
body and therefore the test is misleading.
Change-Id: If430bce75ad5b2fd5ace38d0f65d9c4af56b88df
|
| 9de94e4aaa6de951d1c7ca08d137d57b36d24fa3 |
|
19-Dec-2013 |
Alex Klyubin <klyubin@google.com> |
Fix test breakages due to TLSv1.2 enabled and FS preferred.
Bug: 11220570
Change-Id: Ic918e806ae1d8ea9b98c690cfabd9c3245886525
|
| ec33de3303760e5ba31bd6d0e6b46d5681c123dc |
|
04-Nov-2013 |
Narayan Kamath <narayan@google.com> |
Add a test for lower case auth scheme handling.
bug: 11473660
(Cherry picked from d87ee0f7b587bf19b7bf4c047aa6e5ab0c5045de)
Change-Id: I42fed52e5fdfe0a0ec99dfc8f77330b37a82f772
|
| d87ee0f7b587bf19b7bf4c047aa6e5ab0c5045de |
|
04-Nov-2013 |
Narayan Kamath <narayan@google.com> |
Add a test for lower case auth scheme handling.
bug: 11473660
Change-Id: I21cbff82508921b34b45ce2d62bfc959f6ca4b01
|
| 157b95d87eba99a4804aca480ced9c8b99637dc7 |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update authentication tests.
Test for multiple schemes and multiple auth headers, but
insist that one of schemes is "Basic".
Fixes CTS tests
URLConnectionTest#testAuthenticateWithCommaSeparatedAuthenticationMethods
URLConnectionTest#testAuthenticateWithMultipleAuthenticationHeaders
(cherry picked from d1b438d58c2177c9cebcc3b72df13ac9b7c46e91)
bug: 10211309
Change-Id: Icdc30a5c52cc2c8e3b65abe402732658d4308a88
|
| 0136437c9973615e02f4a22b622399d6a7063932 |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Test that timeouts are respected on recycled connections.
This is a renamed version of testGetKeepAlive, which wasn't
testing very much in its old form.
(cherry picked from 35e58c073df4a45345e89fe8f4d0f77bf7df878)
bug: 10211309
Change-Id: I08c46f577741817b7de714d362f3426d17984042
|
| 8a1e42f2cfcb3cab8cfba7a572d68c534598d627 |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Fix tests dealing with retryable bodies.
Test them in a slightly more robust manner.
- Rely on the fallback from Modern TLS for the retry, instead
of a fallback due to a closed socket.
- Use mode DISCONNECT_AFTER_READING_RESPONSE to make sure the
response is consumed before being discarded.
(cherry picked from 2f1d4ee69ef9dc2eb0d9da37ba4ef1c6089d595e)
bug: 10211309
Change-Id: I8227148e1fc04b7557a41ee6485ffe98efb77ace
|
| d1b438d58c2177c9cebcc3b72df13ac9b7c46e91 |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update authentication tests.
Test for multiple schemes and multiple auth headers, but
insist that one of schemes is "Basic".
Fixes CTS tests
URLConnectionTest#testAuthenticateWithCommaSeparatedAuthenticationMethods
URLConnectionTest#testAuthenticateWithMultipleAuthenticationHeaders
Change-Id: Icdc30a5c52cc2c8e3b65abe402732658d4308a88
|
| 35e58c073df4a45345e89fe8f4d0f77bf7df8787 |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Test that timeouts are respected on recycled connections.
This is a renamed version of testGetKeepAlive, which wasn't
testing very much in its old form.
Change-Id: I08c46f577741817b7de714d362f3426d17984042
|
| 2f1d4ee69ef9dc2eb0d9da37ba4ef1c6089d595e |
|
21-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Fix tests dealing with retryable bodies.
Test them in a slightly more robust manner.
- Rely on the fallback from Modern TLS for the retry, instead
of a fallback due to a closed socket.
- Use mode DISCONNECT_AFTER_READING_RESPONSE to make sure the
response is consumed before being discarded.
Change-Id: I8227148e1fc04b7557a41ee6485ffe98efb77ace
|
| cc440f43ee5cee8b52d6ad86c6926d1189aad37b |
|
20-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Track changes to MockWebServer
Track changes in ag/63871 which add bookkeeping requests
to the request queue.
(cherry picked from b27031a36bbb3c64ad0672553122dcebb17c89)
bug: 10211309
Change-Id: I8093bda336950846dafaaba2acaace8ee52713d6
|
| 3778c428ec5bc08881d8cc8057d79027648b327c |
|
20-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update test exercising broken proxy behaviour.
The test must account for the fact that the connection
will be retried with TLS_MODE_COMPATIBLE if the first
attempt throws an exception.
It's also worth noting that MockWebServer does not
deal with abrupt client disconnections in a particularly
graceful manner (but that does not affect the test,
and I'll address it in a follow up).
(cherry picked from bfdb88fab84756703e01e9b43808a406bdd)
bug: 10211309
Change-Id: I5bf8a8e890df4146ecd0606e50c5a616434fd883
|
| b27031a36bbb3c64ad0672553122dcebb17c89f3 |
|
20-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Track changes to MockWebServer
Track changes in ag/63871 which add bookkeeping requests
to the request queue.
Change-Id: I8093bda336950846dafaaba2acaace8ee52713d6
|
| bfdb88fab84756703e01e9b43808a406bdd48755 |
|
20-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update test exercising broken proxy behaviour.
The test must account for the fact that the connection
will be retried with TLS_MODE_COMPATIBLE if the first
attempt throws an exception.
It's also worth noting that MockWebServer does not
deal with abrupt client disconnections in a particularly
graceful manner (but that does not affect the test,
and I'll address it in a follow up).
Change-Id: I5bf8a8e890df4146ecd0606e50c5a616434fd883
|
| 3d849f3d37a0b7c4c8b7b0fa5a285289ed99a309 |
|
15-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update URLConnection#testLenientUrlToUriNul.
Our earlier behaviour was somewhat broken. We'd accept
hosts with \u0000 in them, and then produce a request with
a bad Host header. We now fail on such requests, so add a
unit test to exercise that new behaviour.
The ideal behaviour might be to be lenient and escape characters
that we don't like. (This would involve using the URI from
URL#toUriLenient to populate the Host header. I attempted that
but it has the side effect of introducing a new set of
incompatibilities).
(cherry picked from aafc934e9e3b5afce7c54162654720fad80)
bug: 10211309
Change-Id: I38a0d39124dc5ca766a6e963e72b0250440e82ee
|
| aafc934e9e3b5afce7c54162654720fad80a733e |
|
15-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Update URLConnection#testLenientUrlToUriNul.
Our earlier behaviour was somewhat broken. We'd accept
hosts with \u0000 in them, and then produce a request with
a bad Host header. We now fail on such requests, so add a
unit test to exercise that new behaviour.
The ideal behaviour might be to be lenient and escape characters
that we don't like. (This would involve using the URI from
URL#toUriLenient to populate the Host header. I attempted that
but it has the side effect of introducing a new set of
incompatibilities).
Change-Id: I38a0d39124dc5ca766a6e963e72b0250440e82ee
|
| 0df27a0c1f64c589a1df3eb2d94c75aea90cdf4a |
|
14-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Fix URLConnectionTest#testConnectTimeouts.
Use StuckServer correctly. When StuckServer isn't
in blocking mode, its local socket address isn't 127.0.0.1.
Attempting to connect to localhost will then throw a
SocketException (ECONNREFUSED) instead of a
SocketTimeoutException.
(cherry picked from commit bf1a9e10eefd28bb3a989df20b4a57c745fb3670)
Bug: 10211309
Change-Id: I8aaab598d043b893e73fab9055082fa420cfb9cd
|
| a37fb255a1993e835eea8c934205b1e568a1271c |
|
14-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Fix URLConnectionTest#testConnectTimeouts.
Use StuckServer correctly. When StuckServer isn't
in blocking mode, its local socket address isn't 127.0.0.1.
Attempting to connect to localhost will then throw a
SocketException (ECONNREFUSED) instead of a
SocketTimeoutException.
(cherry picked from commit bf1a9e10eefd28bb3a989df20b4a57c745fb3670)
Bug: 10211309
Change-Id: I8aaab598d043b893e73fab9055082fa420cfb9cd
|
| bf1a9e10eefd28bb3a989df20b4a57c745fb3670 |
|
14-Aug-2013 |
Narayan Kamath <narayan@google.com> |
Fix URLConnectionTest#testConnectTimeouts.
Use StuckServer correctly. When StuckServer isn't
in blocking mode, its local socket address isn't 127.0.0.1.
Attempting to connect to localhost will then throw a
SocketException (ECONNREFUSED) instead of a
SocketTimeoutException.
Change-Id: I8ea1af80996894a5951e53f9c31e828bbbca3628
|
| 8c213bcc2aadb8249d683312505f1068ae74f1a8 |
|
12-Jun-2013 |
Narayan Kamath <narayan@google.com> |
Fix URLConnectionTest after latest okhttp update.
I've observed 14 failing tests both before and after
the update.
Change-Id: I40dfd9efe35ed03d89b38b030d3c7b6558b247e2
|
| d4781a20611fd0424aede028c139bfdbaea5677b |
|
06-May-2013 |
Brian Carlstrom <bdc@google.com> |
Make it clearer that a new MockWebServer is created for each test run
Change-Id: Ib3ce11f42d19688fe00b7d057eece25567fccea2
|
| 2503556d17b28c7b4e6e514540a77df1627857d0 |
|
30-Mar-2013 |
jwilson <jwilson@squareup.com> |
Switch to OkHttp as URL's preferred HTTP implementation.
Change-Id: Id724b75dd78b68ed00f5db4989c4070896996ec0
|
| 0e1afa1091f74a6228c01d8c7a8eebf001efdc57 |
|
13-Oct-2012 |
Elliott Hughes <enh@google.com> |
Fix ConcurrentCloseTest flakiness.
We can't rely on consuming all the listen(2) backlog. For the tests we've
seen fail because they sometimes connect rather than time out, switch to
an unroutable address.
(cherry-pick of babccbf9e429c4c78aca24c205825ceaaf7d3f37.)
Bug: 6971145
Change-Id: Ibfa412ff1ad7da7e63842d0162cc67a706e2b27e
|
| babccbf9e429c4c78aca24c205825ceaaf7d3f37 |
|
13-Oct-2012 |
Elliott Hughes <enh@google.com> |
Fix ConcurrentCloseTest flakiness.
We can't rely on consuming all the listen(2) backlog. For the tests we've
seen fail because they sometimes connect rather than time out, switch to
an unroutable address.
Bug: 6971145
Change-Id: I259d31b1a15123bcd78c36849d5ed863d392ac20
|
| b11e44a402eb0f9c5da1f7482fc1eabdd9adff47 |
|
20-Sep-2012 |
Jesse Wilson <jwilson@squareup.com> |
Fix a bug in tunnel construction and 'Connection: close' headers.
Based on https://github.com/square/okhttp/pull/30
(git fetch git://github.com/square/okhttp.git && git cherry-pick -x b8c51dc5bdc89d5487cbc110d6fe40389de16480)
Bug: http://code.google.com/p/android/issues/detail?id=37221
Bug: 7291446
(cherry picked from commit 4bedacd569fed2db56c518ed239412b3fe32245e)
Change-Id: I68c55e9259db91dbccca30824c00726094570304
|
| 4bedacd569fed2db56c518ed239412b3fe32245e |
|
20-Sep-2012 |
Jesse Wilson <jwilson@squareup.com> |
Fix a bug in tunnel construction and 'Connection: close' headers.
Based on https://github.com/square/okhttp/pull/30
(git fetch git://github.com/square/okhttp.git && git cherry-pick -x b8c51dc5bdc89d5487cbc110d6fe40389de16480)
Bug: http://code.google.com/p/android/issues/detail?id=37221
Change-Id: I5bed33d9ab16002aa38c2f729fa14a2022485469
|
| 6d41a7cc3cb4cc684c8bece69ddc55954812ad6e |
|
24-Sep-2012 |
Narayan Kamath <narayan@google.com> |
Strip content length in requests with "transparent" gzip handling.
We need to strip both the Content-Length and the Content-Encoding
for such requests. In such requests, it will be the length of the
compressed response. We hide the fact that compression is taking place
from clients, so we shouldn't give them the content length either.
(cherry picked from commit f9d60aed414ae21811a6488f603333517f257b22)
Change-Id: Ic2776ecf020c34b1c55e2fb7e2f0728e553187a0
|
| f9d60aed414ae21811a6488f603333517f257b22 |
|
24-Sep-2012 |
Narayan Kamath <narayan@google.com> |
Strip content length in requests with "transparent" gzip handling.
We need to strip both the Content-Length and the Content-Encoding
for such requests. In such requests, it will be the length of the
compressed response. We hide the fact that compression is taking place
from clients, so we shouldn't give them the content length either.
Change-Id: I80713ab33143945c5e2656f478d83cc9e60226a8
|
| 3edd28a92fc86a1260347d0995e65a815d73bbbe |
|
16-Apr-2012 |
Mattias Petersson <mattias.petersson@sonymobile.com> |
Fix problem with Runtime.runFinalization()
This is a fix for a problem with runFinalization(). The
problem was that all FinalizerReferences to objects that had
not yet been garbage collected were lost when calling this
function. When a FinalizerReference was lost, it is was not
possible to call the finalize() method when the object
was garbage collected.
The result was that finalizers was sometimes never
called, which typically lead to memory leaks.
Also stop synchronizing on the class itself; use a private
lock instead.
Bug: 6907299
Bug: 5462944 # Synchronization on FinalizerReference.class
Change-Id: Ief515edbb5a1823c06d7371415d131165baef7f2
|
| 8155888a2c4e7cc28596deec28433e1984cfaf11 |
|
25-Jul-2012 |
Brian Carlstrom <bdc@google.com> |
Fix HttpURLConnection CloseGuard warning due to lack of close on GZIPInputStream
java.lang.Throwable: Explicit termination method 'end' not called
at dalvik.system.CloseGuard.open(CloseGuard.java:184)
at java.util.zip.Inflater.<init>(Inflater.java:82)
at java.util.zip.GZIPInputStream.<init>(GZIPInputStream.java:96)
at java.util.zip.GZIPInputStream.<init>(GZIPInputStream.java:81)
at libcore.net.http.HttpEngine.initContentStream(HttpEngine.java:523)
at libcore.net.http.HttpEngine.readResponse(HttpEngine.java:831)
at libcore.net.http.HttpURLConnectionImpl.getResponse(HttpURLConnectionImpl.java:274)
at libcore.net.http.HttpURLConnectionImpl.getResponseCode(HttpURLConnectionImpl.java:486)
at ...
Bug: 6602529
Change-Id: I9b49cbca561f8780d08844e566820087fdffc4d7
|
| 4ec2d0dea800397fb32227e0631246f4d2a73191 |
|
29-Jun-2012 |
Brian Carlstrom <bdc@google.com> |
HttpEngine should not attempt to cache proxy CONNECT responses
This fixes an issue where a bad proxy repsonse that included a body
would turn into an IllegalStateException instead of an IOException.
Bug: 6754912
Change-Id: I204ad975820693d6add2780d7b77360463e33710
|
| deb26711801f9bfafcd5cad20963b3c3913404fd |
|
04-Apr-2012 |
Elliott Hughes <enh@google.com> |
Add a test for an already-fixed publicly-reported bug.
Bug: http://code.google.com/p/android/issues/detail?id=28095
Change-Id: I26ff2834becce6f51e668ec71b4e5050b15f06cd
|
| 0731920fdf845358cc13ce78292f9e80e143f915 |
|
28-Mar-2012 |
Brian Carlstrom <bdc@google.com> |
Disable TLSv1.1 and TLSv1.2 by default
Bug: 6234791
Change-Id: I5d829211c9e1d5672fc96e42ef603c53d789e695
|
| 3d74b4bec8543e6e3f89eafe3afe0925f3a69f01 |
|
28-Mar-2012 |
Brian Carlstrom <bdc@google.com> |
Disable TLSv1.1 and TLSv1.2 by default
Bug: 6234791
Change-Id: I5d829211c9e1d5672fc96e42ef603c53d789e695
|
| 3e6dd45baa0d7f9b4fa06f4ade76e088b59cc7bf |
|
16-Mar-2012 |
Brian Carlstrom <bdc@google.com> |
Tracking openssl-1.0.1
Bug: 6168278
Change-Id: I240d2cbc91f616fd486efc5203e2221c9896d90f
|
| df29c7dc69965462cd19de8910b04c8cb463e57f |
|
15-Feb-2012 |
Jesse Wilson <jessewilson@google.com> |
Don't attempt to parse an HTTP response body on HEAD requests.
Previously if the response had a Content-Length or Transfer-Encoding
header we assumed it had a body. This was untrue of HEAD requests.
This is not a regression.
Bug: http://code.google.com/p/android/issues/detail?id=24672
Change-Id: I264db2014e681a6f8e93cd40b0f1640c95103d21
|
| a468471e8ccc2f642d5b2ae3c75873a74267ece2 |
|
26-Jan-2012 |
Jesse Wilson <jessewilson@google.com> |
Don't throw needlessly when returning the CacheResponse.
We were throwing under the premise that we'd always have
response headers before we had a cached body. But when the
user calls connect() we'll have queried the cache but we
won't necessarily have response headers from the server yet.
That's because connect() doesn't require a response to be ready.
Bug: http://b/5920659
Cherry-picked from: I9d69b44b5be3d0e29f14d8b83781f6e65c67582a
Change-Id: Ia403b42ddc9783fd1456164f4b7350da3ce80566
|
| fa5e8dfe3c7ed144b0fbe69091628dedd6a3b961 |
|
21-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Don't reuse an SSL socket if the socket factory has changed.
Bug: http://b/3042192
Change-Id: Iaa9deb69058afcca4ab79ca6ba745c55d1e9321a
|
| abe4e615f473b387b1c00a738062a7c428f05a33 |
|
20-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Unit test SSL fallback.
Bug: http://b/4462288
Change-Id: I9a29c70b24f235df15aa9a763ae5217e10e42f83
|
| 5757cb35ad2e58d7accb7f5d3db6f2d5e72f097d |
|
16-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Support multiple challenges for HTTP authentication.
This fixes the problem reported in the bug, but we still need
to give attention to the digest scheme. Currently we can't do
Basic auth if the HTTP response contains both a Digest challenge
and a Basic challenge!
Bug: http://code.google.com/p/android/issues/detail?id=19081
Change-Id: Ifffb307678c2afa0db1d50c1ef5fc49ec7769306
|
| b2b02ac6cd42a69463fd172531aa1f9b9bb887a8 |
|
15-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Change the way we cope with stale pooled connections.
Previously we'd attempt to detect if a connection was stale by
probing it. This was expensive (it relied on catching a 1-millisecond
read timing out with a SocketTimeoutException), and racy. If the
recycled connection was stale, the application would have to catch
the failure and retry.
The new approach is to try the stale connection directly and to recover
if that connection fails. This is simpler and avoids the isStale
heuristics.
This fixes some flakiness in URLConnectionTest tests like
testServerShutdownOutput and testServerClosesOutput.
Bug: http://b/2974888
Change-Id: I1f1711c0a6855f99e6ff9c348790740117c7ffb9
|
| 2d9fa917aae6a6da38e9d1eda05841ffdf8855bb |
|
15-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Document the effect on timeouts of having both IPv4 and IPv6 addresses
Bug: http://b/3441111
Change-Id: Ib4c2e3e0d8d86eb5eb77cf92a0814c7444fd153f
|
| 0613de89655e481fa610bfd4f1bcaeeae3272205 |
|
14-Dec-2011 |
Jesse Wilson <jessewilson@google.com> |
Document that disconnect() doesn't release sockets.
We could make disconnect() work, but it's a lot of complexity for
little gain. The core problem is that by the time the user calls
disconnect(), the connection has already been pooled. On the RI,
disconnect() will close a socket connection if it hasn't yet been
removed from the pool. Our pooling always-on or always-off behavior
is simpler and more predictable.
Bug: http://b/4148042
Change-Id: I61ef2ccb898f3ffaa08cedc70ee9e728fc1889b9
|
| d0d626655f1d452070d3116678037e8759f807f4 |
|
10-Nov-2011 |
Jesse Wilson <jessewilson@google.com> |
Don't NullPointerException when a URL contains a space.
We end up creating a URI of http://and%20roid which has no host
name part because and%20roid is not a valid host name. It's unclear
where the best place to fail is. This is consistent with the RI
which doesn't fail until it attempts to connect.
Bug: http://code.google.com/p/android/issues/detail?id=16895
Change-Id: Ifa4dc3c651f4627e27622ed5cedfa820f9a35358
|
| 25a753691a80186cd4d7086b12c0e52225d95897 |
|
30-Sep-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix chunked input stream's available() method to never return -1.
Change-Id: I6ea268e469675ef890010282dbe35ed5838cca73
http://code.google.com/p/android/issues/detail?id=20442
|
| 09336c914b4fc813e493acc82469b9ad89fd8694 |
|
07-Jun-2011 |
Jesse Wilson <jessewilson@google.com> |
Use external/mockwebserver in libcore
Change-Id: I4dac34f88b23484643bce31e5f25ac6eb1fea426
|
| 5292410e4ebf7fb5149eefd2f52fcb94c46690a6 |
|
24-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Rewrite parsing for java.net.URL.
This fixes many broken cases on handling relative URLs.
We normalize all URLs by default. This will result
in more URL equality than before. Previously the URLs
http://android.com/a/../ and http://android.com/ were not
equal; now they are equal.
Change-Id: I8cf7be2e42eeb1386520be2698d8f14e0a55decb
http://b/4361656
|
| ec6163ceac64902ccde3259e8a426f0f9ad6a88c |
|
20-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Add test cases inspired by Jetty-dependent Harmony tests.
Change-Id: Id8288e0524b64255c5d9808d3f5582708211d96b
|
| 1e518e80aeb8c27d49d51e5c69173d3e08fb9ec7 |
|
09-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Deduplicate a unit test merged in from upstream.
Change-Id: Ia5f4ad1753fa6bb2ce4dd5fdf4c978e1ff0ffa80
http://b/4188137
|
| 68b8820c82a524388a32501fa9b423d77b4b73b6 |
|
09-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Deduplicate a test that was duplicated by an automerge.
Change-Id: Id8ef04b21c1132db224c69437428bbe4f5890a6a
http://b/4188137
|
| 5e8b5a55eda914b67c62b4e77152922d5c77eb68 |
|
29-Mar-2011 |
Jesse Wilson <jessewilson@google.com> |
Don't read from the delegate stream after we close it.
Honeycomb MR2 patch cherry-picked from bc4c79c6a2059003f695f7ad204de36700e8d701
Change-Id: Ib379595a1e6b2ed3b16d4d82c744b00e4b576c51
http://b/4188137
|
| c996149b500fc4825156106554457fe2394ae087 |
|
06-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix proxy+HTTPS to handle authentication requests from the proxy.
Previously I'd introduced a bug by assuming that CONNECT
responses would always be successful.
Change-Id: Iaf0caf67f52154f6951a20284c75db0090843b99
http://b/4188137
|
| d742d7fc7f0593eb8c6e4ac9dd4c0f6a80374e46 |
|
05-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Use GET after a POST is redirected.
Also dropping support for response code 305 Use Proxy, which
is seldom used, and whose behavior is quite different from
Chrome and Firefox. This is an intentional delta from the RI.
Also testing disconnect().
Change-Id: I80433f4776a2cfc5e228179421a787a44a4ded69
http://b/3180373
|
| 953df613522e12a418cb7cb73248594d6c9f53d4 |
|
03-May-2011 |
Jesse Wilson <jessewilson@google.com> |
Refactor HttpURLConnection implementation.
Break HttpURLConnectionImpl into two parts: the part that
implements the Java API (that follows redirects etc.) and
the part that speaks HTTP. The HTTP part is called HttpEngine,
and it can only be used for a single HTTP request. This makes
it easier to set request properties that only apply to a
single socket request.
Tests for 'Vary', 'Content-Location' and warnings.
For simplicity of implementation I've chosen not to support
'Vary' at this time.
The 'Content-Location' header doesn't require any work from
the cache.
Warnings are ugly but allow the caller to identify when the
cache results may be problematic.
Also breaking HttpResponseCache's dependency on HttpURLConnectionImpl
so it can be tested on the RI.
Change-Id: Idbabb51251f479c2cdea4e0fceb029bfd07182be
http://b/3180373
|
| 0c59055dd24e1659f85d9ff7e2148883f663bd62 |
|
27-Apr-2011 |
Jesse Wilson <jessewilson@google.com> |
Support conditional gets in HttpURLConnection.
This is the first step towards adding an HTTP response cache.
There's still much to do; details are listed on the bug.
Change-Id: Ie693fe424d1d90e90576fc959595e4d96a31f767
http://b/3180373
|
| 5f03a4682a71e56535059258a4d41150b99177d7 |
|
05-Apr-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix a test that was merged from both GB and HC.
Change-Id: I8cf7dec2dcec177669a04951635c95c78d8c7aaf
http://b/4188137
|
| bc4c79c6a2059003f695f7ad204de36700e8d701 |
|
31-Mar-2011 |
Jesse Wilson <jessewilson@google.com> |
Don't read from the delegate stream after we close it.
Change-Id: Ib724b170f7ec10ea0025c0e2e7978119fd070eb6
http://b/4188137
|
| 0c2fd828abec671333b8b88281825fd27a783723 |
|
30-Mar-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix MockWebServer to use the same hostname as TestKeyStore.
Change-Id: I31c9f6e9add87a17737eac4f09949540e5da8cb5
|
| f241d462634527692b7d99335cdc8c11883ac966 |
|
29-Mar-2011 |
Jesse Wilson <jessewilson@google.com> |
Don't read from the delegate stream after we close it.
Change-Id: I9a018ca88373d5f317335e35fc6ca43c5473490e
http://b/4188137
|
| 32559028b14b9b321b10eede050afd554a376569 |
|
09-Feb-2011 |
Jesse Wilson <jessewilson@google.com> |
Use the same code to encode and fix poorly encoded URLs.
Don't permit digits outside of ASCII.
Change-Id: I47b756c32e67b7415e35ae9eb55b735d1c22a36d
http://b/3436051
|
| 38db242d16a5a31fd86c519b817ae94c5fc89417 |
|
27-Jan-2011 |
Elliott Hughes <enh@google.com> |
Consistently use "expected" for expected exceptions in our tests.
Change-Id: I796624fb53dbba420d43858741c9a97b4fb8d017
|
| afd9b157f467b7c4f2f0b5592dca72f18d844602 |
|
20-Jan-2011 |
Jesse Wilson <jessewilson@google.com> |
Handle the CacheResponse returning a null body.
Change-Id: I45c4d6b7cac5cabd8c42f5e17d110fb5b3ecc9cc
http://b/3373699
|
| d1b5e5da828434388e486a388710d21e4306dae0 |
|
20-Jan-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix URL character bugs by adding yet another encoder.
We don't have an encoder that ignores already escaped
characters.
Change-Id: I9ff080ad1745136b7d7313d78af2b0ca1785b200
http://b/3360947
|
| b7f4d6c3968c372767b2510f38a3d506067aced6 |
|
13-Jan-2011 |
Jesse Wilson <jessewilson@google.com> |
Move tests from android.core.URLTest to URLConnectionTest.
This includes some changes to MockWebServer to make it stop
accepting connections as soon as its response queue is empty.
Change-Id: I5a6bcdf1c03d0c36b11552ae086d0dece0440f64
http://b/1158780
|
| fa9f91a486b81b19db13eab59e704179381de22f |
|
12-Jan-2011 |
Jesse Wilson <jessewilson@google.com> |
Fix typo in testResponseCacheCallbackApis().
Embarassing. Test passes with the typo fixed!
Change-Id: Ic348f68c0e720163ca77b125415b01e840c7412d
|
| f14bee4dbf9f724a177a85a7b73f7b0bc6df1ec0 |
|
06-Jan-2011 |
Jesse Wilson <jessewilson@google.com> |
Return HTTP headers even if there is no HTTP response body.
Also add a test to demonstrate that we've bug 7787.
http://code.google.com/p/android/issues/detail?id=7787
Change-Id: I4e320835516c05ec9e305661b5bbb1e59b278183
|
| e942f46f10bb9384a1b186b3d7b74f9704c57090 |
|
17-Dec-2010 |
Jesse Wilson <jessewilson@google.com> |
Test that HttpURLConnection responds gracefully when the server closes a socket.
Change-Id: I5110d58d91e7c6f8f7553ad400f0bd841bcf07b6
http://b/2612240
|
| 2915378e253f08e47fe5a9bfd026cd1ca7c6c351 |
|
16-Dec-2010 |
Brian Carlstrom <bdc@google.com> |
HttpsURLConnection retry should not invoke X509TrustManager and HostnameVerifier more than once
Summary:
In 2.3, HttpsURLConnection was change to retry TLS connections as SSL
connections w/o compression to deal with servers that are TLS
intolerant. However, if the handshake proceeded to the point of
invoking the X509TrustManager, we should not retry. Similarly, if we
should not invoke the HostnameVerifier repeatedly, and need to wait
until the SSL handshake has completed.
Tested with (includes two new tests for this issue):
libcore/luni/src/test/java/libcore/javax/net/ssl/
libcore/luni/src/test/java/libcore/java/net/URLConnectionTest.java
libcore/luni/src/test/java/org/apache/harmony/luni/tests/internal/net/www/protocol/https/HttpsURLConnectionTest.java
Details:
HttpConnection.setupSecureSocket has been broken into two
pieces. setupSecureSocket now just does the SSL
handshaking. verifySecureSocketHostname now does the
verification. The old HttpConnection code was careful never to
assign its sslSocket field until verification was complete. A new
unverifiedSocket field is added to store the sslSocket before
verification is completed by verifySecureSocketHostname.
luni/src/main/java/org/apache/harmony/luni/internal/net/www/protocol/http/HttpConnection.java
HttpsEngine.makeConnection now skips TLS intolerant retry if the
reason for the makeSslConnection failure was a
CertificateException, since that implies that we failed during
certification validation after initial handshaking. We also
prevent retrying hostname verification by moving it out of
makeSslConnection and only doing it on new SSL connections,
tracking the changes to HttpConnection.setupSecureSocket mentioned
above. We also now skip the redundant call to setUpTransportIO in
makeSslConnection on reused SSLSockets.
luni/src/main/java/org/apache/harmony/luni/internal/net/www/protocol/https/HttpsURLConnectionImpl.java
Instead of throwing away the underlying CertificateExceptions, set
them as the cause of the SSLExceptions. This is what the RI does
in the case of X509TrustManager failures and is now used by
HttpsEngine.makeConnection.
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLSessionImpl.java
luni/src/main/java/org/apache/harmony/xnet/provider/jsse/OpenSSLSocketImpl.java
Added new testConnectViaHttpsToUntrustedServer which makes sure
that connections are not retried on certificate verification
failure.
luni/src/test/java/libcore/java/net/URLConnectionTest.java
Added new test_SSLSocket_untrustedServer that verifies that an
SSLHandshakeException is thown containing a CertificateException
is thrown on certificate verification problems.
luni/src/test/java/libcore/javax/net/ssl/SSLSocketTest.java
Added second test CA and a new TestKeyStore.getClientCA2 test key
store that does not trust the primary test key stores. This is
useful for negative testing and is used in the above two new
tests.
support/src/test/java/libcore/java/security/TestKeyStore.java
Issue: http://code.google.com/p/android/issues/detail?id=13178
Bug: 3292412
Change-Id: I37136bb65f04d2bceaf2f32f542d6432c8b76ad4
|
| c6dae581716b9362a5c7f166c80a7f2b46ed1124 |
|
06-Nov-2010 |
Jesse Wilson <jessewilson@google.com> |
Throw the expected exception after the response failed.
Change-Id: I4c2ed726d41d6b55e27a503467a412fa05242383
http://b/3172197
|
| 35eef71e8ce721c4199c525890ecc1a263054596 |
|
02-Dec-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix two bugs with HTTPS + file uploads.
Firstly, we weren't echoing calls to setFixedLengthStreamingMode
and setChunkedStreamingMode from the HTTPS adapter down to the
real HTTP client.
Secondly, we were calling setTransportIO twice, and the second
call destroyed the buffer of our request body stream. That buffer
has only existed since change Id22a8449118a4b8dd13e71915700ac78803d2d9f,
(November 2, 2010) so the impact of that problem is limited.
Change-Id: Icb0eb2e0ca4e201a629bd74c3b98e852f6cc0fd3
http://code.google.com/p/android/issues/detail?id=12860
|
| 253851cdb2904082c205949da273c455d197c083 |
|
01-Dec-2010 |
Jesse Wilson <jessewilson@google.com> |
Do not merge: Fix a bug where we weren't pooling connections due to mismatched URLs.
Already committed to master as change 78504.
Change-Id: Ib883e92bb03b142f34449ef1314c57878dcc00f0
http://b/3169861
|
| dbc42ad94ce37b3178e14b50496914c267d6cea6 |
|
16-Nov-2010 |
Brian Carlstrom <bdc@google.com> |
HttpsURLConnection fixes to pass URLConnectionTest and HttpsURLConnectionTest
Add public getSecureSocket so that HttpsURLConnection can reset its sslSocket field on reused connections.
luni/src/main/java/org/apache/harmony/luni/internal/net/www/protocol/http/HttpConnection.java
discardIntermediateResponse now maintain old values for
intermediateReponse, which is necessary when retrying the CONNECT
method with proxy authorization.
luni/src/main/java/org/apache/harmony/luni/internal/net/www/protocol/http/HttpURLConnectionImpl.java
Clear sslSocket before retrying makeSslConnection to ensure we reconnect.
makeSslConnection now resets sslSocket on resumed connection.
makeSslConnection now exits early on existing connection.
luni/src/main/java/org/apache/harmony/luni/internal/net/www/protocol/https/HttpsURLConnectionImpl.java
git cherry-pick -e 4df5be29bdffc09e7368c03423a0bef85fa26ac5
Bug: 3184701
Change-Id: Ida3c027f79e5e29968263ac761d4f4f79d063a27
|
| 5bcbe1c84468a25fec1e5b4e1dd636f917c73422 |
|
16-Nov-2010 |
Brian Carlstrom <bdc@google.com> |
HttpsURLConnectionTest rewrite and URLConnectionTest updates
Rewrote HttpsURLConnectionTest to work with current SSLSocket
implementation which more strictly follows RI behavior.
Also made upates to URLConnectionTest related to HttpsURLConnection
- changed assertContent call connect() explictly to illustrate problem
with https proxy case also seen by HttpsURLConnectionTest
- Rewrote testConnectTimeouts to work reliably. Before if often
worked on WiFi but not on mobile networks where a "HTTP/1.1 501 Bad
Gateway" would be seen causing a FileNotFoundException.
- Changed testConnectViaHttpProxyToHttpsUsingHttpProxySystemProperty
to match expectation that https does not use the http.proxyHost
values. Added new testConnectViaHttpProxyToHttpsUsingProxyArgWithNoProxy
which should have the same behavior as the fixed
testConnectViaHttpProxyToHttpsUsingHttpProxySystemProperty
git cherry-pick -e f02c695ed03e708623d9365dec26d533356ef2d0
Bug: 3184701
Change-Id: Id25f619d2437db607deaf35aeb1d5e817514b92f
|
| 5fc5dde4c719c1dfdac46b67d5d2e4884d07721e |
|
16-Nov-2010 |
Elliott Hughes <enh@google.com> |
Improve ConcurrentCloseTest.
No more flaky use of 10.* addresses.
Bug: 3044772
Change-Id: I5ca8dc431b50950efdc818efe73eb9aba76ea67f
|
| f02c695ed03e708623d9365dec26d533356ef2d0 |
|
16-Nov-2010 |
Brian Carlstrom <bdc@google.com> |
HttpsURLConnectionTest rewrite and URLConnectionTest updates
Rewrote HttpsURLConnectionTest to work with current SSLSocket
implementation which more strictly follows RI behavior.
Also made upates to URLConnectionTest related to HttpsURLConnection
- changed assertContent call connect() explictly to illustrate problem
with https proxy case also seen by HttpsURLConnectionTest
- Rewrote testConnectTimeouts to work reliably. Before if often
worked on WiFi but not on mobile networks where a "HTTP/1.1 501 Bad
Gateway" would be seen causing a FileNotFoundException.
- Changed testConnectViaHttpProxyToHttpsUsingHttpProxySystemProperty
to match expectation that https does not use the http.proxyHost
values. Added new testConnectViaHttpProxyToHttpsUsingProxyArgWithNoProxy
which should have the same behavior as the fixed
testConnectViaHttpProxyToHttpsUsingHttpProxySystemProperty
Change-Id: I397f55905524e2bbb587533f258ef0418ca8af80
|
| 06e15e6c528fcb773bedb43e34b0577312570927 |
|
05-Nov-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix a bug where we weren't pooling connections due to mismatched URLs.
Change-Id: I13bebc94189192c3f8d959e521989591d2fca3e2
|
| c1a675c80c69decadb736b245f0366f93a94a462 |
|
29-Oct-2010 |
Jesse Wilson <jessewilson@google.com> |
Remove the tree map in the HTTP headers.
This map is intended make looking up headers faster, but for real-world
sized datasets we spend more energy maintaining the map than we save from
using it.
I benchmarked the Android Market's HTTP connections' header use and it's
significantly faster to skip the map:
run us linear runtime
map+list 413.1 ==============================
list only 65.8 ====
Change-Id: I5b8fef6f999eb1bfe53eea2781f63cdd1002a273
|
| 8ac847a52e72f0cefbb20a6850ae04468d433a9e |
|
28-Oct-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix broken interactions between HTTP response caches and redirects.
This fixes several specific problems:
- HTTP header map didn't contain the status line (under the null key)
- response code, message and version weren't set by caches
- caches didn't work with redirects
This change also makes some significant cleanup to the Headers class.
We might be able to simplify this further by dropping the map, should
that prove efficient enough.
Change-Id: Ib79ec17bef5978b3234f68102114eee7d4b7cda2
http://b/3139211
http://b/3139211
|
| 37dcf5581f177229ca6c8e7d0d640361640bfb00 |
|
27-Oct-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix ResponseCache to support caching of HTTPS responses.
Previously it would fail with an internal error because of
inconsistencies because HttpURLConnectionImpl claimed to be
'connected' when it had a cache hit, and HttpsURLConnection
acted upon this by talking to its sockets.
Change-Id: I51f4215ceb9c5fd851223a501488306fa6d382b1
http://b/3043966
|
| 984fcff696380abd6ea14e80030f9fd2d09dbad0 |
|
14-Oct-2010 |
Jesse Wilson <jessewilson@google.com> |
Tests for HTTP bugs using HTTPS through a proxy.
Change-Id: I01b470204d4da46016794ab7040d5751e7116958
http://b/3097277
http://b/3097518
|
| f29ad8a60254345d1943d1b3836482395a7c916f |
|
30-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Don't explode if flush() is called when an HTTP upload stream is closed.
Streams like BufferedOutputStream always call flush() even if it is
a no-op. These flushes must be permitted for compatibility with apps
that don't call close() until after the response has been retrieved.
Change-Id: Ic5c86ab6050c6c4d166c44ae5b4fc7a1688e7e45
http://b/3038470
|
| 1f8243e3d2b5a3f8e0398c304d1dea0395cbc368 |
|
28-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Remove java.net tests that are redundant with Harmony.
Change-Id: I01d6b3633a6de8a075982975d4cbf6c08e757490
|
| 00feece22909b7dc79fc96d666d157390b93858e |
|
24-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Strip usage of the term 'localhost' from URLConnectionTest.
Change-Id: I8ea7923c3ea72728c7df64c13bdd6f94b91be121
http://b/3032912
|
| ef66494dce45a0b7ec22ec3fb20c60096517a4e3 |
|
22-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix problem where single-byte reads were unsigned.
Change-Id: Ib0bc273698b71d13a90a03a8c60498ad7de5ad9d
http://b/3023872
|
| d4bddd7d1fb7b1b7f0836648228235c6e4b56a18 |
|
16-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix two HTTP issues that came up when writing HttpURLConnection docs.
We're now more careful about which headers are sent to HTTP proxies.
And getContentEncoding() is better documented.
Change-Id: I04241f99c2f32c25ba005fbd6ff9ef7236c3c9d3
|
| 8116f7e97e00d223e7fbe5c950c9a5e3277de124 |
|
17-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Don't leave the Content-Encoding header around after transparent gzip.
Otherwise clients may be tempted to double-decompress.
http://b/issue?id=3009828
Change-Id: I4832da1c2aff9bad8d452ffc4a0f98ee27d44f49
|
| 65d890eb22aeba9b009ee642ffd4fff48a6f98ae |
|
08-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Cope with HTTP responses where the response code and headers disagree.
See http://b/issue?id=2981779
Change-Id: I0e4ed8f0cc606aa419704ba2eb5cd9938b5ff320
|
| b7155fd57239e986bbaba254a91aeb9600d60305 |
|
03-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Fixing a broken pipe with HttpsURLConnection connection reuse.
We had a bug where we were peeking a byte from the raw socket
stream rather than the encrypted stream when we were attempting
to reuse an HTTPS connection. The raw stream field shouldn't
have been initialized, but setupTransportIO() did it while the
connection was being established.
This fixes setupTransportIO() to not initialize the fields
and isStale() to use the SSL fields if they exist.
See http://code.google.com/p/android/issues/detail?id=8625
Change-Id: Idd5b6f20e098adf436997829940707548896bedb
|
| 50ae32218918eae80298bd1ab8e4f588bbbabdb2 |
|
01-Sep-2010 |
Jesse Wilson <jessewilson@google.com> |
Configure HttpURLConnection to prefer gzipped content by default.
This is not what the RI does. But the bandwidth savings is substantial,
and we can offer it almost for free.
See http://code.google.com/p/android/issues/detail?id=9390
Change-Id: I43c3fae8938f33809f2526d2d0e1cc5e08ae1202
|
| f418bf447fd007cd2ec2d45b4b0399a11904e9b4 |
|
30-Aug-2010 |
Jesse Wilson <jessewilson@google.com> |
Support "Connection: close" from both request and response.
This is a rewrite of fix for the same problem in AOSP master:
https://review.source.android.com/16757
This expands on that fix by supporting "Connection: close" on
redirects by adding test coverage.
Change-Id: Ia0daa9d83cf9f557b4e9a99b2c51345c98c6db14
|
| 125f068f0a6cd739beac97821c9421cf8317cc87 |
|
20-Aug-2010 |
Jesse Wilson <jessewilson@google.com> |
Fix a bug when user sets "Content-Encoding: chunked" manually.
We used to try to build a -1-sized ByteArrayOutputStream, which
failed miserably with an exception.
Change-Id: Ia6fa72c639a14c03f03b6f73083ce37ffab8b4e5
|
| 4557728efb66c455a52b7669a8eefef7a9e54854 |
|
11-Aug-2010 |
Jesse Wilson <jessewilson@google.com> |
Moving tests to be under the libcore.* package.
This is indended to make it easier to run on VMs that restrict the packages
from which application classes can be loaded. For example, on the RI you need
to use the bootclasspath to load these tests.
Change-Id: I52193f35c5fcca18b5a3e1d280505b1e29b388af
|