Build boot images that prefer Zedboot over local boot.
Default value: false
Directory containing signing keys used by amber-publish.
Default value: "//garnet/go/src/amber/keys"
Default value: "//root_build_dir/amber-files/repository/blobs"
Directory containing files named by their merkleroot content IDs in ASCII hex. The //build/image:amber_publish_blobs target populates this with copies of build products, but never removes old files.
Default value: "//root_build_dir/amber-files"
Default value for target_cpu = "arm64": false
Default value for target_cpu = "x64": true
This is a list of targets that will be built as vulkan ICDS. If more than one target is given then use_vulkan_loader_for_tests must be set to true, as otherwise tests won't know which libvulkan to use.
Default value for target_cpu = "arm64": []
Default value for target_cpu = "x64": ["//third_party/mesa:magma_vulkan"]
Default value for target_cpu = "arm64": true
Default value for target_cpu = "x64": false
Default value for target_cpu = "arm64": true
Default value for target_cpu = "x64": false
Default value: "../buildtools/linux-x64/clang/bin"
Determines various flavors of build configuration, and which concrete targets to use for dependencies. Valid values are “standalone”, “chromium”, and “fuchsia”. Defaulted to “fuchsia” because “is_fuchsia_tree” is set.
Default value: "fuchsia"
Default value: ""
Default value: ""
The size of the minfs data partition image to create. Normally this image is added to FVM, and can therefore expand as needed. It must be at least 10mb (the default) in order to be succesfully initialized.
Default value: "10m"
Default value: true
Default value: true
Default value: false
Default value: true
Default value: "//third_party/expat"
Additional SSH authorized_keys file to include in the build. For example: extra_authorized_keys_file="$HOME/.ssh/id_rsa.pub"
Default value: ""
Additional variant toolchain configs to support. This is just added to known_variants, which see.
Default value: []
List of packages (a GN list of strings). If unset, guessed based on which layer is found in the //.jiri_manifest file.
Default value: [[]](https://fuchsia.googlesource.com/build/+/master/gn/packages.gni#8)
Current value: ["garnet/packages/buildbot"]
The size in bytes of the FVM partition image to create. Normally this is computed to be just large enough to fit the blob and data images. The default value is "", which means to size based on inputs. Specifying a size that is too small will result in build failure.
Default value: ""
Default value: "//third_party/glm"
Absolute directory containing the Goma source code.
Default value: "/home/swarming/goma"
Default value: "undefined"
Default value: "x64"
Default value: "linux"
This is the directory where host tools intended for manual use by developers get installed. It‘s something a developer might put into their shell’s $PATH. Host tools that are just needed as part of the build do not get copied here. This directory is only for things that are generally useful for testing or debugging or whatnot outside of the GN build itself. These are only installed by an explicit install_host_tools() rule (see //build/host.gni).
Default value: "//root_build_dir/tools"
Tells icu to load an external data file rather than rely on the icudata being linked directly into the binary.
This flag is a bit confusing. As of this writing, icu.gyp set the value to 0 but common.gypi sets the value to 1 for most platforms (and the 1 takes precedence).
TODO(GYP) We'll probably need to enhance this logic to set the value to true or false in similar circumstances.
Default value: true
Debug build.
Default value: true
File containing kernel command line arguments to roll into the bootdata image used for booting.
Default value: ""
List of variants that will form the basis for variant toolchains. To make use of a variant, set select_variant (which see).
Normally this is not set as a build argument, but it serves to document the available set of variants. See also universal_variants. Only set this to remove all the default variants here. To add more, set extra_variants instead.
Each element of the list is one variant, which is a scope defining:
configs (optional) [list of labels] Each label names a config that will be automatically used by every target built in this variant. For each config ${label}, there must also be a target ${label}_deps, which each target built in this variant will automatically depend on. The variant() template is the recommended way to define a config and its _deps target at the same time.
remove_common_configs (optional) remove_shared_configs (optional) [list of labels] This list will be removed (with -=) from the default_common_binary_configs list (or the default_shared_library_configs list, respectively) after all other defaults (and this variant's configs) have been added.
deps (optional) [list of labels] Added to the deps of every target linked in this variant (as well as the automatic “${label}_deps” for each label in configs).
name (required if configs is omitted) [string] Name of the variant as used in select_variant elements' variant fields. It's a good idea to make it something concise and meaningful when seen as e.g. part of a directory name under $root_build_dir. If name is omitted, configs must be nonempty and the simple names (not the full label, just the part after all /s and :s) of these configs will be used in toolchain names (each prefixed by a “-”), so the list of config names forming each variant must be unique among the lists in known_variants + extra_variants.
toolchain_args (optional) [scope] Each variable defined in this scope overrides a build argument in the toolchain context of this variant.
host_only (optional) target_only (optional) [scope] This scope can contain any of the fields above. These values are used only for host or target, respectively. Any fields included here should not also be in the outer scope.
Default value: [{ configs = ["//build/config/lto"] }, { configs = ["//build/config/lto:thinlto"] }, { configs = ["//build/config/scudo"] }, { configs = ["//build/config/sanitizers:ubsan"] }, { configs = ["//build/config/sanitizers:ubsan", "//build/config/sanitizers:sancov"] }, { configs = ["//build/config/sanitizers:asan"] host_only = { remove_shared_configs = ["//build/config:symbol_no_undefined"] } }, { configs = ["//build/config/sanitizers:asan", "//build/config/sanitizers:sancov"] host_only = { remove_shared_configs = ["//build/config:symbol_no_undefined"] } }, { name = "asan_no_detect_leaks" toolchain_args = { asan_default_options = "detect_leaks=0" } configs = ["//build/config/sanitizers:asan"] host_only = { remove_shared_configs = ["//build/config:symbol_no_undefined"] } }]
Default value: "//garnet/lib/magma"
Enable this to have the msd include a suite of tests and invoke them automatically when the driver starts.
Default value: false
Enable this to include fuchsia tracing capability
Default value: true
Default value: "/b/s/w/ir/kitchen-workdir/third_party/mako"
Default value for target_cpu = "x64": "//third_party/mesa"
No values for target_cpu = "arm64".
Enable all 8 cores, which is faster but emits more heat.
Default value for target_cpu = "arm64": true
No values for target_cpu = "x64".
With this flag set the system tries to use cache coherent memory if the GPU supports it.
Default value for target_cpu = "arm64": true
No values for target_cpu = "x64".
Default value for target_cpu = "x64": false
No values for target_cpu = "arm64".
Default value: "//garnet/drivers/gpu/msd-intel-gen"
Default value: ""
Sets the default LTO type for rustc bulids.
Default value: "unset"
Sets a custom base directory for rustc and cargo. This can be used to test custom Rust toolchains.
Default value: "//buildtools/linux-x64/rust/bin"
Default value for target_cpu = "arm64": 1
Default value for target_cpu = "x64": 0
Default Scudo options (before the SCUDO_OPTIONS environment variable is read at runtime). NOTE: This affects only components using the scudo variant (see GN build argument select_variant), and does not affect anything when the use_scudo build flag is set instead.
Default value: ["abort_on_error=1", "QuarantineSizeKb=0", "ThreadLocalQuarantineSizeKb=0"]
The directories to search for parts of the SDK.
By default, we search the public directories for the various layers. In the future, we'll search a pre-built SDK as well.
Default value: ["//garnet/public", "//peridot/public", "//topaz/public"]
List of “selectors” to request variant builds of certain targets. Each selector specifies matching criteria and a chosen variant. The first selector in the list to match a given target determines which variant is used for that target.
Each selector is either a string or a scope. A shortcut selector is a string; it gets expanded to a full selector. A full selector is a scope, described below.
A string selector can match a name in select_variant_shortcuts, which see. If it‘s not a specific shortcut listed there, then it can be the name of any variant described in known_variants and universal_variants (and combinations thereof). A selector that’s a simple variant name selects for every binary built in the target toolchain: { host=false variant=selector }.
If a string selector contains a slash, then it‘s "shortcut/filename" and selects only the binary in the target toolchain whose output_name matches "filename", i.e. it adds output_name=["filename"] to each selector scope that the shortcut’s name alone would yield.
The scope that forms a full selector defines some of these:
variant (required)
[string or false] The variant that applies if this selector
matches. This can be false to choose no variant, or a
string that names the variant. See `known_variants` and
`universal_variants`.
The rest below are matching criteria. All are optional. The selector matches if and only if all of its criteria match. If none of these is defined, then the selector always matches.
The first selector in the list to match wins and then the rest of the list is ignored. So construct more complex rules by using a “blacklist” selector with variant=false before a catch-all or “whitelist” selector that names a variant.
Each “[strings]” criterion is a list of strings, and the criterion is satisfied if any of the strings matches against the candidate string.
host
[boolean] If true, the selector matches in the host toolchain.
If false, the selector matches in the target toolchain.
testonly
[boolean] If true, the selector matches targets with testonly=true.
If false, the selector matches in targets without testonly=true.
target_type
[strings]: "executable", "loadable_module", or "driver_module"
output_name
[strings]: target's `output_name` (default: its `target name`)
label
[strings]: target's full label with `:` (without toolchain suffix)
name
[strings]: target's simple name (label after last `/` or `:`)
dir
[strings]: target's label directory (`//dir` for `//dir:name`).
Default value: []
This should never be set as a build argument. It exists only to be set in toolchain_args. See //build/toolchain/clang_toolchain.gni for details.
Default value: []
List of short names for commonly-used variant selectors. Normally this is not set as a build argument, but it serves to document the available set of short-cut names for variant selectors. Each element of this list is a scope where .name is the short name and .select_variant is a a list that can be spliced into select_variant, which see.
Default value: [{ select_variant = [{ variant = "asan_no_detect_leaks" host = true dir = ["//third_party/yasm", "//third_party/vboot_reference", "//garnet/tools/vboot_reference"] }, { variant = "asan" host = true }] name = "host_asan" }]
List of extra packages to synthesize on the fly. This is only for things that do not appear normally in the source tree. Synthesized packages can contain build artifacts only if they already exist in some part of the build. They can contain arbitrary verbatim files. Synthesized packages can't express dependencies on other packages.
Each element of this list is a scope that is very much like the body of a package() template invocation (see //build/package.gni). That scope must set name to the string naming the package, as would be the name in the package() target written in a GN file. This must be unique among all package names.
Default value: []
The package key to use for signing Fuchsia packages made by the package() template (and the system_image packge). If this doesn‘t exist yet when it’s needed, it will be generated. New keys can be generated with the pm -k FILE genkey host command.
Default value: "//build/development.key"
Default value: ""
Current value for target_cpu = "arm64": "arm64"
Current value for target_cpu = "x64": "x64"
Default value: ""
The absolute path of the sysroot that is used with the target toolchain.
Default value: ""
ThinLTO cache directory path.
Default value for target_cpu = "arm64": "host_x64/thinlto-cache"
Default value for target_cpu = "x64": "thinlto-cache"
Number of parallel ThinLTO jobs.
Default value: 8
Manifest files describing target libraries from toolchains. Can be either // source paths or absolute system paths.
Default value for target_cpu = "arm64": ["/b/s/w/ir/kitchen-workdir/buildtools/linux-x64/clang/lib/aarch64-fuchsia.manifest"]
Default value for target_cpu = "x64": ["/b/s/w/ir/kitchen-workdir/buildtools/linux-x64/clang/lib/x86_64-fuchsia.manifest"]
This should never be set as a build argument. It exists only to be set in toolchain_args. See //build/toolchain/clang_toolchain.gni for details. This variable is a scope giving details about the current toolchain: toolchain_variant.base [label] The “base” toolchain for this variant, often the right thing to use in comparisons, not current_toolchain. This is the toolchain actually referenced directly in GN source code. If the current toolchain is not shlib_toolchain or a variant toolchain, this is the same as current_toolchain. In one of those derivative toolchains, this is the toolchain the GN code probably thought it was in. This is the right thing to use in a test like toolchain_variant.base == target_toolchain, rather rather than comparing against current_toolchain. toolchain_variant.name [string] The name of this variant, as used in variant fields in select_variant clauses. In the base toolchain and its shlib_toolchain, this is "". toolchain_variant.suffix [string] This is “-${toolchain_variant.name}”, "" if name is empty. toolchain_variant.is_pic_default [bool] This is true in shlib_toolchain. The other fields are the variant's effects as defined in known_variants.
Default value for target_cpu = "arm64": { base = "//build/toolchain/fuchsia:arm64" }
Default value for target_cpu = "x64": { base = "//build/toolchain/fuchsia:x64" }
Default value: [{ toolchain_args = { is_debug = false } configs = [] name = "release" }]
Default value: true
Set to true to enable compiling with ccache
Default value: false
Set to true to enable distributed compilation using Goma.
Default value: false
Use link time optimization (LTO).
Default value: false
Default value for target_cpu = "x64": false
No values for target_cpu = "arm64".
Enable the Scudo memory allocator.
Default value: false
Use ThinLTO variant of LTO if use_lto = true.
Default value: true
Mesa doesn‘t properly handle loader-less operation; their GetInstanceProcAddr implementation returns 0 for some interfaces. On ARM there may be multiple libvulkan_arms, so they can’t all be linked to.
Default value: true
Default value: "warn,error"
File containing kernel command line arguments to roll into the bootdata image used for zedboot.
Default value: ""
Default value: true
Manifest files describing extra libraries from a Zircon build not included in zircon_boot_manifests, such as an ASan build. Can be either // source paths or absolute system paths.
Since Zircon manifest files are relative to a Zircon source directory rather than to the directory containing the manifest, these are assumed to reside in a build directory that's a direct subdirectory of the Zircon source directory and thus their contents can be taken as relative to get_path_info(entry, "dir") + "/..". TODO(mcgrathr): Make Zircon manifests self-relative too and then merge this and toolchain_manifests into generic aux_manifests.
Default value for target_cpu = "arm64": ["//out/build-zircon/build-arm64-ulib/bootfs.manifest"]
Default value for target_cpu = "x64": ["//out/build-zircon/build-x64-ulib/bootfs.manifest"]
Groups to include from the Zircon /boot manifest into /boot. This is either “all” or a comma-separated list of one or more of: core -- necessary to boot misc -- utilities in /bin test -- test binaries in /bin and /test
Default value: "core"
Manifest files describing files to go into the /boot filesystem. Can be either // source paths or absolute system paths. zircon_boot_groups controls which files are actually selected.
Since Zircon manifest files are relative to a Zircon source directory rather than to the directory containing the manifest, these are assumed to reside in a build directory that's a direct subdirectory of the Zircon source directory and thus their contents can be taken as relative to get_path_info(entry, "dir") + "/..".
Default value for target_cpu = "arm64": ["//out/build-zircon/build-arm64/bootfs.manifest"]
Default value for target_cpu = "x64": ["//out/build-zircon/build-x64/bootfs.manifest"]
Zircon build directory for target_cpu, containing .manifest and .bin files for Zircon's BOOTFS, BOOTDATA, and kernel image.
Default value for target_cpu = "arm64": "//out/build-zircon/build-arm64"
Default value for target_cpu = "x64": "//out/build-zircon/build-x64"
Default value: "//zircon"
TODO(mcgrathr): Could default to "" for !is_debug, or “production build”. Note including “test” here places all of Zircon's tests into /system/test, which means that Fuchsia bots run those tests too.
Default value: "misc,test"
Where to find Zircon's host-side tools that are run as part of the build.
Default value: "//out/build-zircon/tools"