- Mar 18, 2019
-
-
Todd Eisenberger authored
When all of a composite device's components are ready, the composite is created. In this implementation, we always proxy components, even if they're in the same process. This will be relaxed in a later patch. Issue: ZX-2648 Change-Id: I97f5efb311ca18ec830f81158f022f230abc2d5d
-
James Tucker authored
system/bin only contains appmgr, which is launched by full path. /pkg/bin is not relevant to the shell, the shell should launch through the /bin stubs. Change-Id: Idd079b12d2afb346a8f4f5dfda7a045c1dbc865d
-
Eric Wang authored
Test: Mechnical change Change-Id: I4a32c5e9f178b2a7202fe0dccecc54661f0198cc
-
Kostya Kortchinsky authored
The standalone version of Scudo is currently being upstreamed in LLVM's compiler-rt, this is meant to provide a temporary way for Fuchsia to use it as a replacement for jemalloc in musl. Some of the files have already been reviewed and commited upstream, while others haven't, and are likely to undergo some changes during the review process. The LLVM copyright header is present in all files, even the ones that are not yet upstream, which I am not sure is legitimate. BUILD.gn was provided by David. Note that hardware CRC32 is not enabled. Test: replace jemalloc with scudo in musl, fx full-build & fx run. Change-Id: I42d6dbefd25e7c2252a18f712243b72b73b04ed3
-
Nina Chen authored
- Include symptoms that indicate troubleshooting is necessary - Split troubleshooting steps into subsections - Add to subsection about resolving pub errors Change-Id: I46260e30e4376448c0b3a5ff579dc2245d3b2202
-
Didrik Nordström authored
Integrate AccountEventEmitter into AccountManager, which lets external clients subscribe to account changes (such as when an account is added or removed) using the AccountListener FIDL interface. AUTH-144 Tested=CQ Change-Id: I82ab51bb6ee54329015b3ea23f7564d53c7f6b0d
-
P.Y. Laligand authored
Bug: DX-447 Change-Id: I817c263ec4a3e0d9864e89ac77ee2b607168949c
-
Gianfranco Valentino authored
Added myself and brettw as owners of the library. TEST=none Change-Id: If1a7e7f4e941f015bc4b425ba336e54dd631cdd4
-
Christopher Anderson authored
Some PCI devices like serial consoles and framebuffers are brought online by the BIOS/bootloader and have existing BAR allocations. For BAR bringup this was not maintained in the userspace driver, but for continued work moving forward it was necessary. If the bus driver sees something that could be a valid address in the address bits of a given device BAR then it will attempt to allocate that address from the upstream allocator, if and only if the device has been enabled prior to our probing. Unfortunately, at power-on the state of all address bits are indeterminate, so the best we can do is preserve any address we see present and then attempt to use it. If we can allocate it then we don't need to change anything and we can proceed. If not, then we allocate a properly sized region from upstream and continue with BAR configuration. This should match the existing behavior in the kernel PCI bus driver. ZX-3146 Change-Id: I974d903b9d2a4e99069b874a1a2c0389c9e50630
-
Dale Sather authored
The ffmpeg demux node in mediaplayer occasionally gets a request (from downstream) for a packet even when it has reported end-of-stream on all streams. This happens only rarely. https://fuchsia-review.googlesource.com/c/fuchsia/+/263095 addressed the first-encountered DCHECK. That just got us to the next DCHECK. This CL should prevent this failure in CQ going forward. MTWN-242 #done TEST: fx shell run_test_component mediaplayer_tests \ --gtest_repeat=<many> # debug build Change-Id: I0b070760621d0f8b78124be44aa28a5358b3c2a7
-
Rob Tsuk authored
This takes the matching boilerplate out of the apps. Modified the button sample to respect focus and be closer to Material Design. Change-Id: I76c4f1b95592a1f45abe731580521cf159e0933e
-
Benjamin Brittain authored
Change-Id: If58b0c800f79eb6dff5d6bd014723f120cd66233
-
Gianfranco Valentino authored
static SetUpTestCase/TearDown must be public. TEST=zxtest-integration-test Change-Id: I6ba055b801728e817d6e59ea4b8c0298335b783c
-
Adam Barth authored
Supporting both the old monolith/preinstall/available names and the new base/cache/universe names will make it easier to transition to the new names. Change-Id: Ifad1ce4250ab8683a1786e49882c89b43447b5a9
-
Nathaniel Manista authored
Change-Id: Ia7cd416d103a47ec84293ac60a25f4214cb49d1e
-
Benjamin Lerman authored
When an asynchronous method M is called on an object O, and O is deleted before the callback is called, the callback should not be called later. This fixes a couple of instance where the callback could be called after the source object has been deleted. Change-Id: Iebe3a2a86fd27cab1149567e24192ff34630ee58
-
Vikram Auradkar authored
O(logn) alloc and free time complexities. Use IdAllocator to alloc and free blobfs inodes. The existing code uses O(n). Not only allocation take a lot of time but it destroys the cpu caches because it is walking array of blobfs_inode_t structure. Around 128KiB should be sufficient for (1<<20) Ids. Tested: Ran blobfs-test. Added id-alloc-bitmap test cases. Performance: Test created 10000 blobs. In a 10000 iteration loop, 1. unlink first created and last created blobs 2. create two new blobs. Creating two blobs would stress FindNode. Total time(ns) spent in FindNode on Pixelbook, before: 768521344 after : 45631811 FindNode on Qemu, before: 6220833923 after : 691256103 Change-Id: I3f2619b65c8a63cd8374a541ad3ec1ad385a6bf6
-
James Tucker authored
Change-Id: I17ead26f066be9bc18f4b207acfb5ff43334be48
-
Francois Rousseau authored
* we need that service to get feedback data for crash analysis (part of core) and user feedback (up the stack) * while this increases the core image, this will be compensated by the removal of several Dart agents up the stack * also moving crashpad_agent from //garnet/packages/product:base to //products/core DX-957 #comment TESTED=`fx set x64 --product core` && `fx shell crasher` Change-Id: I16e226aa3bdf0c3664be48f53e6590dd6338f0bb
-
Craig Tiller authored
Change-Id: I67bc1e5503fa0852e2212ce3f52bd711ac612acb
-
Dale Sather authored
The prebuilt artifacts for the ffmpeg lib are now present at //prebuilt/third_party/ffmpeg. This CL moves the BUILD.gn used for ffmpeg clients to //src/media/lib/ffmpeg. References in //garnet/bin/mediaplayer and //garnet/bin/mediasession are updated to depend on //src/media/lib/ffmpeg. The .gitignore in garnet/lib/media/ffmpeg will be removed once the prebuilts no longer land there. TEST: no behavior change Change-Id: Ic304cc3ebfba78a8b283103d261e27bc674dd26d
-
Benjamin Lerman authored
The following status are converted from one to the other: OK, IO_ERROR, PAGE_NOT_FOUND, KEY_NOT_FOUND, REFERENCE_NOT_FOUND, NETWORK_ERROR Any other status will be converted to an INTERNAL_ERROR. Change-Id: I36020de7f72eee4784c840a208d2becb5def0b7c
-
Sean Klein authored
ZX-2948 #comment In Progress Change-Id: I4b9b963f7ab364d5a0b32c32681545aba3d7b963
-
Adam Barth authored
Improves the error message when the caller does not provide a PRODUCT.BOARD goal. Change-Id: I1343510e0dfe0938e4c05e7403948129c1e68717
-
Benjamin Lerman authored
Document that the page must exist before trying to call |EvictPage|. Change-Id: I29322f971fefd0daffd159a4601459e7c028e955
-
Zach Bush authored
Change-Id: Ibc4fe4b606b2c35aa229141389bd851cd24a9ee7
-
Sean Klein authored
Change-Id: Idca9c6d478cede154b7ef568e2101941073c8915
-
Brian Bosak authored
This patch moves the usb-virtual-bus into its own devmgr, and adds a test binding for it. It also removes usb-virtual-bus functionality from usbctl, as this will be tested in an automated fashion in a future CL. In order to control the virtual bus, it is necessary to start an isolated devmgr and create a bindable device with BIND_PLATFORM_DEV_VID set to PDEV_VID_TEST and BIND_PLATFORM_DEV_PID set to PDEV_PID_USB_VBUS_TEST. Then; ioctls can be used to manipulate the virtual bus. Test: Validate that usbctl no longer contains any virtual bus functionality, and that the virtual bus is no longer visible in the device tree. Change-Id: Ia3ae6eb89c71ecafae95b427b1cb0659da206fb5
-
Gabriel Kerneis authored
Introducing a distinct type name will ease refactorings. Switching to a set makes the pathological split test case fast again. A follow-up CL will switch to a map to distinguish lazy and eager values. Test: fx run-test ledger_tests -t ledger_unittest Bug: LE-697 Change-Id: Ifa225f2462ac89b6feb7ee8fcfb2ae77f6b7672a
-
Przemyslaw Pietrzkiewicz authored
With this patch, Voila reaches a first functional version: - we now create two isolated replicas and run session shell in each - Ledger instances in replicas synchronize via cloud_provider_in_memory LE-663 Test: tiles_ctl add fuchsia-pkg://fuchsia.com/voila#meta/voila.cmx Change-Id: I5f8c889fa383c7d9731f071e563705307430f2cd
-
Adam Barth authored
Change-Id: I03463d8a3528f4c31972f678924cd7a643c67d7a
-
Przemyslaw Pietrzkiewicz authored
In a subsequent CL, we'll add similar support to fuchsia-component. LE-663 Test: see https://fuchsia-review.googlesource.com/c/fuchsia/+/263612 Change-Id: I14678357e2fba7e51fea27699f88b83ee52f0d20
-
Adam Barth authored
This CL starts the process of replacing the JSON build packages with GN group declarations. This CL makes it possible to have a mix fo JSON and GN declarations and migrates many of the JSON declarations in this repository to GN. Change-Id: I0ea3fce91da8a289c30d02e7692ee0f7c4d033c1
-
Nina Chen authored
- fix broken markdown links by removing the space between the [link text] and (#link url) - Flutter can run on "platforms", not "platform" :) - link "Channels" as [Channels] and not [Channel]s so that screen readers read it as "channels" instead of "channel, s" - put bullets on their own line so they render correctly as bullets Change-Id: I7178320f8f70b89edcc811521a5e379c503d2b0e
-
- Mar 17, 2019
-
-
Yifei Teng authored
When blobs are streamed via gzip, their original content length is lost. This broke fetching of packages. TEST: Fetched and ran fortune. Change-Id: I85bbaa255f4e0fed28a29d603869f578aba63588
-
Joshua Seaton authored
This file is meant to be a supplment to tests.json that provides information about the latter's complement, namely tests that don't yet contibute test spec metadata. copy_host_tests is made to consult this manifest as well as tests.json Bug: IN-819 Change-Id: Idc6906365b2dfc9d77d20bcf6598b9901fd712b9 Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/264247 Reviewed-by:
Adam Barth <abarth@google.com> Commit-Queue: Joshua Seaton <joshuaseaton@google.com>
-
Rob Krten authored
Added xunions description and diagram to the wire format doc. Change-Id: Icd3bec41683ce47133f468b4f88d5527dec8179d
-
Adam Barth authored
Rather than generating individual copy rules, we now use tests.json and a Python script to copy host tests into the host_tests directory. After this change, the host tests machinery can be driven entirely off GN metadata. Eventually, we will remove the host_tests directory. IN-819 #comment Change-Id: I02e1b7d7e7fd84cf19ce145ae9559d67a4ca721e
-
Adam Barth authored
Also, make config-data into a meta package. A meta package is part of the monolith that depends on all the other packages in the universe. We need to do some fancy dependency tricks to avoid circular dependencies. Change-Id: I667dca887dc00262717391ddbe2a2f5f9306c3c9
-
- Mar 16, 2019
-
-
Adam Barth authored
People missing meta_package_labels are most likely not using a supported product configuration. This assert can detect that condition early and loudly. Change-Id: Ic5e7b3db0ef7cde278cc25e7c4153fc7666f4484
-