Fix crash in WV_Teardown() causes by deleting shared output parser object.
We used to have a single output parser object and a single memory chunk object
that are shared by multiple sessions. When we delete the output parser, another
session may still be using it. The deletion causes a native crash intermittently.
The fix addresses these two issues: i)output parser object is now created per
session, and ii)add reference count to keep track of the single memory chunk
object use so we do not delet it while other sessions are still using it.
Merge of https://widevine-internal-review.googlesource.com/#/c/9253/ and
https://widevine-internal-review.googlesource.com/#/c/9271/ from widevine repo
RC/4.5.0/601_AndroidOS_K branch.
Change-Id: I6426b544a6d0c264ec91a89e92153862b87607f3
related-to-bug: 13208359
Some make and model information was not being reported correctly to the
license server. "HTC One X+" was encoded as "HTC+One+X+" instead of
"HTC+One+X%2B", which when decoded becomes "HTC One X ". Make and model
information is now correctly url encoded.
Merge of https://widevine-internal-review.googlesource.com/#/c/9240/
from the widevine repo.
b/12722013
Change-Id: I987691c850648f92b408f8749327ff43b13b43a7
This change updates gMock to the new release, 1.7.0. This is
necessary for Android, as Android Master has updated their gTest to
1.7.0, and we must always use the matching version of gMock.
This CL cleans up some removed files that were missed in the
previous commit.
Change-Id: I67ab17425aea8d44fff765f79737b6eb23d16b52
This change updates gMock to the new release, 1.7.0. This is
necessary for Android, as Android Master has updated their gTest to
1.7.0, and we must always use the matching version of gMock.
This should not break any existing tests, as 1.7.0 is
backwards-compatible with 1.6.0 code in nearly all cases. There are
a few bugfixes around being too generous with type coercion in
EXPECT_THAT() and ASSERT_THAT() that could break code that was
accepted by the compiler before but was never technically safe.
For a full list of changes, including all the awesome new matchers
you can now use in your tests, see CHANGES, which is included from
gMock unchanged.
For a full list of modifications made to allow this to work on
Android Master, see the updated README.android.
No changes to the GYP files were necessary as part of this upgrade.
Change-Id: Ib1445044e78c9fe0cf16031d544577d65ebbf6df
A bug (b/8132682) in "Download as PDF" from Google Docs makes table of
contents links point back to the original document. This causes
problems with external customers.
This change uses the following work around: print the document, and
select "Save as PDF" as the printer.
This only removes the links. It does not fix them. When bug 8132682
is fixed we can re-download the documents with proper links.
bug: 12585419
Merge of https://widevine-internal-review.googlesource.com/#/c/8775/
from the widevine cdm repo.
Change-Id: I5344b85bee213aba3a99913a4bb7be824fdd7acf
Support builds for multiple architectures the same time:
Use LOCAL_MODULE_TARGET_ARCH to restrict building to specific arches
Use LOCAL_C_INCLUDES_x86 to set the include path for only x86 builds
Change-Id: I7c33c27f1c9bfb6e3318a07514698992482f6cd7
A bug (8132682) in "Download as PDF" from Google Docs makes table of
contents links point back to the original document. This causes
problems with external customers.
This change uses the following work around: print the document, and
select "Save as PDF" as the printer.
This only removes the links. It does not fix them. When bug 8132682
is fixed we can re-download the documents with proper links.
bug: 12585419
Change-Id: I4e4b0ff8d083192fd2db6a569d1c3f84c8ec4fa9
bug: 12228689
If the device ID returned from OEMCrypto_GetDeviceUniqueId is
not NULL terminated in the OEM code, trailing garbage characters
may be included in the license request's client_identification
field, which could be rejected by the server's utf8 parser if
they are invalid characters, causing a license request failure.
The code for CryptoSession::GetDeviceUniqueId should use the
updated id_length from OEMCrypto_GetDeviceUniqueId to adjust
the length of the *device_id string before returning the result
to the caller.
Change-Id: I659866d4234d4f21ec051590fc7bc6367904a48a
During session sharing, when a sample contains both clear and encrypted
subsamples, subsample flags would on occasion be set incorrectly. Clear
subsamples would be sent to the current session, while encrypted ones
would incur a key id to session lookup and be sent to the appropriate session.
The sessions would then receive decrypt calls with subsample flags
incorrectly set.
In order for this to work correctly all subsamples within a sample need to be
sent to the same session. This requires that key ids be specified and
checked if at least one of the subsamples is encrypted. If however none of
the subsamples are encrypted then a valid key id may not have been provided
to MediaCrypto, and the subsamples may be sent to any session.
In order to support this, the CDM decrypt will now allow the caller to
specify whether to validate the key Id.
Then a check is added to wvcrypto determine whether to ask the CDM to
validate the key ID based on the clear/encrypted states of the subsamples.
The list of subsamples is already being preprocessed, so this
additional check just determines if any subsamples are encrypted, and sets
the validation flag appropriately.
b/11967440
Merge of https://widevine-internal-review.googlesource.com/#/c/8510/3 and
https://widevine-internal-review.googlesource.com/#/c/8520/2 from the
widevine cdm repo.
Change-Id: If65c36a31e56b69f514f0cc547a0becf0c54c40a
