- Oct 02, 2017
-
-
Doug Evans authored
Change-Id: I0532be24f15feeb3c5ef32199311756d66f596d7
-
Doug Evans authored
TO-127 #comment patch Change-Id: I5c0fe34181326a79c91f24e3c24b07e71f86c1eb
-
Doug Evans authored
TO-127 #comment patch Change-Id: I8ac9efe7fead736d9bfa81bd4225ea219d079d7d
-
Doug Evans authored
TO-127 #comment patch Change-Id: I3effa6048f63327de2a70cb4ae1028cb9a9914fa
-
Doug Evans authored
TO-127 #comment patch Change-Id: I15e04380e7077d352e4bef7dfbda561d82d90fc4
-
Doug Evans authored
TO-127 #comment patch Change-Id: I103df1f85cbd6958b1aca54191692d8980ccdc0d
-
John Bauman authored
It only makes sense to allow deps on mesa when the Intel GPU driver is in use, so check for deps in other places so they don't accidentally sneak in. Also do the same thing with the arm mali driver. Change-Id: Id995467893251aab607dc70660c984bae31b1d38
-
Dale Sather authored
The video renderer view needs to register for a callback, which is basically just set_connection_error_handler for the view listener. It deletes itself in that callback. This fixes a memory leak in media_service. The view holds a reference to a VideoFrameSource, which integrates a VideoConverter, which has a multi-megabyte lookup table. Change-Id: I7578f130d868450e9e5ac403380ca23871c42fad
-
John Bauman authored
These need to use PlatformDevice instead of PlatformPciDevice there. Change-Id: I5376cf9906510a806e2414a9eeab4156b619dc06
-
Dmitry Vyukov authored
Currently prints: connect: addr={%!s(tcpip.NICID=0) fe80::6cc2:f3ff:fedb:2a8c %!s(uint16=50123)}, <nil> Change-Id: Ie840d9a5c55785f036822b8ebd50556309896265 Signed-off-by:
Dmitry Vyukov <dvyukov@google.com>
-
Taylor Cramer authored
FIDL MessageResponse Futures register their task to be notified when a response arrives. Prior to this change, this registration would only occur the first time the MessageResponse was polled. However, if a MessageResponse future was moved between tasks, the wrong task would be awoken. To prevent this, this commit stores the last-registered task in the MessageResponse and uses this to detect if the current task has changed so it can reregister. Change-Id: Id80f1f94a2179c1fcbbb34f2e2a6cbd4a864510e
-
Andrew Walbran authored
Change-Id: I832adc5bc501115e3972e1fc2dc02eaf9fc1b511
-
Andrew Walbran authored
Change-Id: Ia159080bf7081f0f058a243070029beec89eecd9
-
Andrew Walbran authored
Change-Id: I9e78fd340545caccf4b0463c4e7f6811a8c1e924
-
Andrew Walbran authored
Change-Id: I09f9022d0e7ab7e7c01eb47665b2f0c030e7e76c
-
Andrew Walbran authored
Change-Id: Ib025819d7fbd6afd66843ffdf2ff12ac883cc037
-
Andrew Walbran authored
Change-Id: I81b5dfa1822acfe1774257a4f943d21fffa8d43b
-
Andrew Walbran authored
Change-Id: Ibf9e3120735bf1edc953f21ad3b80ddce3ffa150
-
Anthony Stange authored
up as a driver. Also, shifts some stuff around in ddk-rs. Change-Id: I31f7c4909279932ec67d521e4c03896722748466
-
Craig Stout authored
On some platforms, display and gpu are provided by different drivers. Unimplement release callbacks since these are never called. Change-Id: Ic13e7ab10798a603e10c9b3f209fcc389c3047c5
-
- Sep 30, 2017
-
-
Taylor Cramer authored
While FIDL clients can reasonably return errors to be handled by the user, FIDL services don't currently have an out-of-band way to return protocol errors or other failure messages. When a server is unable to process a request but a response is required, the only way to indicate failure to the client is to close the channel. When no response is required, the server can choose to either swallow the error or close the channel. Change-Id: I244e11d351c2ad130d582354632e75897634c9ab
-
Taylor Cramer authored
Change-Id: I0a47ab9907cbf8f703536ebc01e8805a6c781957
-
- Sep 29, 2017
-
-
Ross Wang authored
media_packet_producer_base has a couple log statements that may be compiled away. A mitigation was already present for one of the affected captures. Extending it to the other. Change-Id: Iab8486dd8bde54ebf3d79da77138b361c2d6a3d4
-
P.Y. Laligand authored
Change-Id: I8c30a5ce5fa11fc630034bbaae78217f675750ce
-
P.Y. Laligand authored
Change-Id: Ie6d7ce700debc2d7c02fcedd2f02fb2df4d37ae9
-
Josh Gargus authored
Change-Id: I8cdd05aad4527232cb1ededdcd700798e246a0f0
-
P.Y. Laligand authored
Change-Id: I12a663f35c6d863a2dbb7552b25642a850bd1de5
-
P.Y. Laligand authored
Change-Id: I0272b8bf82b90292af0dbdc105e8770031cab659
-
James Tucker authored
Change-Id: Ia2ca0f1984cd7aa68f423c1374f9e87d6db6f1d1
-
P.Y. Laligand authored
Change-Id: I20bace63e7f48690b811af7bbe15ab1418b6b34e
-
Michael Jurka authored
Change-Id: Ie77e25003adb770d9a13d63e3106bb71df912f32
-
Craig Stout authored
Change-Id: Ib95f9e8f380e591b99fde3c66aade62aa4b7df99
-
Taylor Cramer authored
Change-Id: I4601dfe9352a5480c315a77e3a39d4e70c9e2d61
-
Erik Gilling authored
This is needed for process names. Change-Id: I2bcb7195639410c82f42092d51d54729c3cdbdd9
-
Taylor Cramer authored
Change-Id: Id5fb5a09d884aae6d9f8d8dcf7a420406fcbd08c
-
Przemyslaw Pietrzkiewicz authored
Change-Id: I301187550b0d0053498a4d11fcc3ef8d6c8863cf
-
- Sep 28, 2017
-
-
Taylor Cramer authored
Create the fidl-service crate to manage common tasks for Rust FIDL services. Change-Id: I7f04d66c126721b9f0f7dd46a107f3ea7c6126a5
-
Martin Puryear authored
Initial implementation of the media_client library contained exhaustive parameter checks, striving to make the library crash-proof. Realistically, the library runs in-proc, so error checks only defend apps against themselves -- relatively low value. This change converts any remaining NULL-checks to DCHECKs, and eliminates the "bad-handle" checks on audio_manager and audio_output_stream (previously done upon every API call). One bonus: rough perf profiliing shows an approx. improvement of 100 nsec per API call. More concretely, having less code is usually a good thing. Change-Id: I6e2e947e2512f3f84775d39aeb2c79334fc41dfa
-
John Bauman authored
They're only built when build_libvulkan_arm_mali is set. The tests' dependencies on mesa are also removed. Change-Id: Ia08dbb5d254690845eb8144e7639d01d2c05c357
-
John Bauman authored
Allows libvulkan to create and import buffers. Change-Id: Id515b7a2c7b7c73e2422b04997d34d9f333213f7
-