- Apr 06, 2019
-
-
Roland McGrath authored
Nothing uses the Zircon board lists via GN or .../zircon-boards.list any more. Bug: BLD-437 #comment Removed //zircon/public/config Change-Id: Ic9f59756aa95cb43699f1ba926f674620eb34ff1
-
Roland McGrath authored
Update some stale documentation to refer to the Zircon GN build. Mention that //zircon/public is a transitional part of an ongoing migration to a unified GN build. Update the OWNERS files to require review by the people directly involved in that ongoing migration. Bug: BLD-353 Change-Id: Ie212fd48e4dec816ba15dc656fcd6dff1a41eb9b
-
Petr Hosek authored
Always use libc++ that's shipped as part of our Clang toolchain rather than relying on the system verion similarly to what we do on Linux. Due to difference in driver behavoir on Darwin, this requires manually passing the list of static libraries to the linker. Change-Id: I0e8c6a678e8c6de9a0535ce529715d054c46051f
-
Alex Min authored
factory. - Code forks in basemgr_impl, where it will use session_user_provider_impl when account manager is enabled, and user_provider_impl when it's not - SessionUserProviderImpl implements UserProvider API's with account manager to support existing login flows. AddUser() and Login() are the ones to look at. - Guest login will not create an account, and generate token managers from token manager factory. TMF will be removed once ephemeral accounts are supported. MF-226 #done TEST: fx run-tests modular_tests with flag enabled, manual testing of login flows Change-Id: I8535d4c2b5e05b09deb086e2a237c0210e4156b4
-
Jinwoo Lee authored
Change-Id: I5c4e76934723caba23d58db70aef9a3bc29aa993
-
Eric Holland authored
Add ShutdownHook in order to disable hardware. This is being done to prevent bad hardware state after an mexec. DdkSuspend sent to SimpleAudioStream will result in ShutdownHook being called. TEST: tested on two most recent revisions of Astro Change-Id: I04821885937260d414c4ab4ba75d88ff5919ac01
-
Eric Holland authored
Added support to SimpleAudioStream so that Shutdown will be called when DdkSuspend is received. This helps avoid unknown hardware state when mexec is used. TEST: tested on recent two revisions of Astro hardware. Change-Id: I2be7650b89082d3297422a80b89913717ba0bcc6
-
Yifei Teng authored
FIDL-245 #done Mechanical conversion from fidl::Array to std::array TEST: CQ Change-Id: I92a8cf565f9fbca4e41f2c7151170271f6dc92f3
-
Ross Wang authored
For lists in GN, it is common to start with [] and append. However, an empty args list is still defined, so it was attaching an empty option to the script invocation. Adding a gate on empty args. Test: Coded a dart_test target to have args = [] Change-Id: I8ba60eac04c0933d221788379ef41d085a803dc3
-
ankitdave authored
As part of replacing old SemanticsRoot with new Semantics Manager, this cl is adding SemanticsManager and SemanticTree entries in services.config file. Testing: * Build: Passed * Installed image on device. * Unit Test: Unit test for this cl are added in https://fuchsia-review.googlesource.com/c/fuchsia/+/265270 MI4-1736 #comment Change-Id: I59aa7d8ee290c42533bc94ffb4a00bb451950d09
-
Rich Kadel authored
Moves command_line_parser from src/developer/debug/zxdb into zircon/system/ulib under a new "cmdline" directory. With the namespace change, zxdb::CommandLineParser is now renamed to cmdline::ArgsParser. Unit test was copied to "utest" and updated to work with "unittest". Tests pass. I tried to make the minimum changes to the original source, in this CL, with plans for some enhancements in future CLs. The one change I did make (besides changing the namespace from zxdb to cmdline) was to remove the dependency on zxdb::Err. I assume the Err class is used in many more places, compared to ArgsParser, so I'm hoping it's OK to leave zxdb::Err where it is. Within ArgsParser, I replaced zxdb::Err with cmdline::Status, a new class generally based on the zxdb::Err design. zxdb::Err also had a dependency on a printf-style message formatter. I decided to forgo the variadic argument list feature, and just pass a std::string for now, since this is _currently_ only populated by ArgsParser. For readability, I made some small tweaks to the API so the library user can now call like: return cmdline::Status::Ok() or return cmdline::Status::Error("error message") Test: cmdline-test Change-Id: Ia03d4840c43a9df545f9b7af33fe296412709439
-
Jeremy Manson authored
Structs could previously be nullable, but they couldn't hold the value null. Change-Id: I27e22730b6419da2fcf228c7ea0717db4aa02dcc
-
John Grossman authored
Change-Id: I80e84d4cb96fe8d7c06431ebba5ff7b675dd0a97
-
Didrik Nordström authored
Given an auth context provider, and an auth provider type, attempt at provisioning a new account. This may or may not require input from the user. The typical case involves the user entering their credentials in a UI context. This can fail for many different reasons, such as network, authentication, or internal issues. Tested: CQ AUTH-75 Change-Id: I399edca1d4a3a742a698d8edf474b67146fcac2f
-
Oliver Hunt authored
A simple library providing a single API to retrieve a the current time over an HTTPS connection. Bug: CF-701 This requires some contortions during trust evaluation, because evaluation requires the current time: * First we blindly trust the certificates we receive during the handshake, and then parse out date from the http date header. * We then verify the date is at least slightly plausible by requiring it be after the system build date. * Finally we perform a true trust evaluation against the certificates we received during the handshake, using the response date. Obviously this is less than ideal, but it is the best we can do given the constraint that we don't already know the current date and time. By validating against the system rootstore, and verifying that the response date is after the build date, we've at least blocked expired certs, and requiring that the server cert was issued by a real CA. There is no revocation checking, as it is unclear if that would have true value given the limited trust we have in the entire connection. In order to achieve the deferred certificate validation it was necessary for us to opt in to the "dangerous_configuration" feature of rustls. Our tests do hit the network, but consider network failures to be valid results. Change-Id: If71f4153d3c7b7bc5c3f1f01749093ecbbfeccc3
-
Robert Lord authored
It turns out Flutter expects selection and composition range indices to be specified in UTF-16 code units. Since Rust strings are UTF-8, it seems impractical to have the unit throughout the code be these code units, not to mention the existing code already expects bytes. This change adds a conversion step any time a legacy TextInputState is either sent in from a InputMethodEditorClient, or we send a DidUpdateState to a client, which allows the clients to speak these UTF-16 code units, and us to internally continue to use bytes. Optimally, we'd refactor out the byte usage and only use code points, but this seems like overkill right now considering many parts of ime.rs will be deleted after the migration to IME API v2. This change also refactors out some byte index -> code point index conversion that was happening in the proxy to the new TextField interface, which uses code points. (To make the distinction clear — a code unit is a two-byte chunk in UTF-16, analogous to bytes in UTF-8. One or two code units in UTF-16, or one to four bytes in UTF-8, comprise a Unicode code *point*.) A couple of tests in ime.rs have been changed to reflect the index unit change. I've also modified one test for an out-of-range selection that I believe had incorrect behavior. Change-Id: I11cc6689658952f5d0a43e5cceb4ad6ab58d17a0
-
Travis Geiselbrecht authored
In qemu 3.1 some of the device tree nodes we're parsing moved from 'foo' to 'foo@number' syntax. Tweak our device tree node matching code to handle this. Tested: boot on various arm qemu versions to make sure we still work. Change-Id: Ic22e879ae79bd4b37a0ce53d5304364648b79234
-
ankitdave authored
Previously, I have added support to log semantic tree. This cl is building on top of it to expose this semantic tree log through a debug file. There will be 1 file per view in debug directory. Testing: * Build: Passed * Installed image on device. * Unit Test: Unit test for this cl are added in https://fuchsia-review.googlesource.com/c/fuchsia/+/265270 MI4-1828 #comment Change-Id: I358aaee41f9dfb23750527e841f6d0c11f023adc
-
- Apr 05, 2019
-
-
Cristián Donoso authored
The handling was getting messy, so reorganized the test_data dependencies into a group. TEST=No functional change. Change-Id: I7c79fbac2e0b746ace108c4810edf391b155dfe7
-
Taylor Cramer authored
Change-Id: I7db379afd748a3bf86dfd62cc670ceedc13cbdba
-
Allan MacKinnon authored
Change-Id: I939cb52868f04e2c9f5c66e28b22ac8ab269448f
-
Roland McGrath authored
This makes it easy to convert an old ZBI to a new compression algorithm or level setting. Change-Id: Ia5e50e6ec597289fa002c7e6340faf12aeb61040
-
Didrik Nordström authored
- Add a flag --dev-auth-providers to the AccountManager binary which sets the available auth providers to a set consisting only of non-networked "dev auth providers", hard-coded to return fixed responses for testing purposes. - Build and include `dev_auth_providers` as a standalone package in the testing group instead of manually including it in BUILDs for each user. - Correct BUILD and cmx files required for dev_auth_provider. - In integration tests for AccountManager, pass the abovementioned flag. - NOTE: The auth provider code path is not yet invoked by existing tests, but will be in a future CL. Change-Id: I35f33883563238dd5804ffd2ab89f401916cd81d
-
Francois Rousseau authored
DX-957 #done #comment TESTED=`fx run-test crashpad_agent_tests` TESTED=`fx shell crasher` (report id a55f80b3c6def696) TESTED=`fx shell k crash` (report id 08466e192d52aeda) Change-Id: If60f93c21f2bf8dd64ab2ddf7fc5257f9052d0ce
-
Dave Schuyler authored
This CL adds the async loop in the Harvester main() and a Harvester class. Change-Id: I83680403d5418d62995aae70b91496af915b0d89
-
Kostya Kortchinsky authored
Updating the code with upstream reviewed code and some performance improvements. Test: fx run Change-Id: Iaddd5bf7a3f436fc5863f4c8be9f3e8fef544668
-
P.Y. Laligand authored
This version identifier will be used to communicate breaking schema changes to consumers. Bug: DX-1277 #done Change-Id: I080941b6a6272f7b4931f91beab12f74b86390ab
-
Louis Yung-chieh Lo authored
For those missed in provious CLs. TESTED=Compiled in local and 'fx format-code --all' BUG=WLAN-895 Change-Id: I93d7fc142a4d9afdc624117cb5d4c5df9bd637b0
-
John Grossman authored
Update the way in which bits are allocated in handles, and formalize the guarantees we provide moving forward. Prior to this, user-mode encodings of handles were 32 bits, and had the following guarantees. 1) The value 0 always indicated an invalid handle. 2) The LSB of the handle would always be 1. 3) The MSB of the handle would always be 0. Unfortunately, #1 was the only documented guarantee. #2 and #3 were undocumented, but #3 had been implied a very long time ago when we used to use non-negative values as handle return types, and negative values as error codes. The practice of mixing zx_handle_t and zx_status_t ended a long time ago, however. Still, these guarantees are important for some code; specifically for things like futex based mutex implementations which take advantage of the spare bits in order to record mutex ownship as well as contested state in their atomic futex storage. Now, the guarantees have been updated and are... 1) The value 0 will indicate an invalid handle. 2) The two LSBs of the handle will always be 1. Note: Because of the nature of the kernel -> user mode handle obfuscation, it is not possible to guarantee that a user mode handle will have any of its top 30 MSBs set. In particular, the value 0x00000003 can absolutely be a valid user-mode handle encoding. These guarantees have now been formally stated in the zircon documentation which discusses zx_handle_t. Also, update MUSL's pthread_mutex, R/W lock, and condvar to align with the new packing. Prior to this, there were a large number of places where the pthread mutex code made assumptions about the structure of the handle bits. Assumptions were both implicit (such as comparing (state < 0) to determine if a handle/contested state was contested) and explicit (such as or'ing the literal 0x80000000 into a handle in order to encoded an owned-and-contested mutex state). So; this CL also makes an attempt to clean this up a bit. It centralizes all of the state manipulations in static inline functions, and removes all of the explicit/implicit assumptions about state. Finally, it aligns all of the state manipulations with the current definition of handle bits outlined above. Tests: runtests on x64/arm64 QEMU, Intel NUC and VIM2 Note: in order to make this more of a soft transition, we secretly are continuing to force the MSB of handle to be 0 for a little while longer. _No one_ should be depending on this. As soon as all external code making this assumption has been updated, we will remove this. Change-Id: I943a19388f823b7d3678da52377af531a364fce0
-
John Grossman authored
Change-Id: Ic34b2cb1e25ef2c2375420d07be8fa6872d1f90e
-
Roland McGrath authored
Bug: ZX-1751 Change-Id: I317a2a5cf4dbc1557f809231af7aedc60d0123eb
-
David Gilhooley authored
Add a README to easily show how the hello_world component is built, run, and tested. Change-Id: I6c563cdf10f9bbc67fe44259ad515066c0a8c7f9
-
John Bauman authored
This lets us check whether it can be imported. MA-576 #comment Change-Id: Ic7bb491f979a0bc480330e5477c6d4cb52efb5c0
-
Mukesh Agrawal authored
When constructing a StartOfSdu header, we should set the SAR bits to indicate that this is the first segment. Make it so. Bug: BT-442 Test: bt-host-unittests Change-Id: I8655e9c131d098fe3f0761b153f8b4482688a28b
-
Ariana Mott authored
Follow-up from comments on fxr/264530. Change-Id: I92dae2f6a5364103cf2fdff7101566bd1f768a72
-
David Pursell authored
Adds support for job and process exception channels: * expand syscall support for job/process * refactor shared code into the Exceptionate * expand testing to cover the newly supported cases Does not yet support synthetic exceptions (e.g. thread start/stop); this will be added next. ZX-3072 Test: `k ut exception` and `runtests -t exception-test` Change-Id: I98973af6817f1ca3d946a3276900e59147702fcf
-
Craig Tiller authored
Change-Id: I8d1e9e903828855d9f2d61a8531807db3f9a11be
-
John Bauman authored
This was hitting an assert now. Change-Id: I8ad9885ebb1325e8c06fa868746f8e07138f1622
-
Robert Lord authored
Also updates all the comments in the legacy FIDL to be doc comments. Change-Id: I89ce2a15c873f6a55aea54080bfb77e023f86c58
-
Craig Tiller authored
Change-Id: Ifd2eb290637a5c2e7c72f68afb14754592d0b77b
-