skia2/bazel/cc_binary_with_flags.bzl
Kevin Lubick f3726c61d2 [infra] Use Bazel transitions to allow cc_binary to set their own flags
This will hopefully let us pre-package certain binaries (e.g. DM,
fuzz) with more sensible defaults and not make the developer
type out all the settings.

For CanvasKit, which specifies its own build flags, I think I'll
need to make another transition setup, which would go in something
like modules/canvaskit/ck_binary_with_flags.bzl or something.

Some sausage-case-names were converted to snake_case_names as per
go/build-style#target-naming

The example this is based off is worth a look through before
diving into this:
https://github.com/bazelbuild/examples/tree/main/rules/starlark_configurations/cc_binary_selectable_copts


Change-Id: Ia919d47f4d1aa25cf294af7918e36d38838c179e
Bug: skia:12541
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/472688
Reviewed-by: Ben Wagner <bungeman@google.com>
Reviewed-by: Leandro Lovisolo <lovisolo@google.com>
2021-11-18 19:18:54 +00:00

130 lines
5.5 KiB
Python

"""
This file contains a way to set flags from BUILD.bazel instead of requiring users to set them from
the CLI.
It is based off of https://github.com/bazelbuild/examples/tree/main/rules/starlark_configurations/cc_binary_selectable_copts
"""
_string_flags = [
"//bazel/common_config_settings:fontmgr_factory",
"//bazel/common_config_settings:with_gl_standard",
]
_string_list_flags = [
"//bazel/common_config_settings:gpu_backend",
"//bazel/common_config_settings:include_decoder",
"//bazel/common_config_settings:include_encoder",
"//bazel/common_config_settings:include_fontmgr",
]
# These are the flags that we support setting via set_flags
_flags = _string_flags + _string_list_flags
def _flag_transition_impl(settings, attr):
rv = {}
for key in settings:
# Get the short form of the name. This the short form used as the keys in the
# set_flags dictionary.
flag_name = key.split(":")[1]
# If there is an entry in set_flags for the short-version of a flag, use that
# value or values. If not, use whatever value is set via flags.
flag_setting = attr.set_flags.get(flag_name, settings[key])
if key in _string_list_flags:
if type(flag_setting) == "list":
rv[key] = flag_setting
else:
rv[key] = [flag_setting] # This usually happens when the default value is used.
elif key in _string_flags:
if type(flag_setting) == "list":
rv[key] = flag_setting[0]
else:
rv[key] = flag_setting # we know flag_setting is a string (e.g. the default).
return rv
# This defines a Starlark transition and which flags it reads and writes.
_flag_transition = transition(
implementation = _flag_transition_impl,
inputs = _flags,
outputs = _flags,
)
# The implementation of transition_rule: all this does is copy the cc_binary's output to
# its own output and propagate its runfiles and executable to use for "$ bazel run".
#
# This makes transition_rule as close to a pure wrapper of cc_binary as possible.
def _transition_rule_impl(ctx):
actual_binary = ctx.attr.actual_binary[0]
outfile = ctx.actions.declare_file(ctx.label.name)
cc_binary_outfile = actual_binary[DefaultInfo].files.to_list()[0]
ctx.actions.run_shell(
inputs = [cc_binary_outfile],
outputs = [outfile],
command = "cp %s %s" % (cc_binary_outfile.path, outfile.path),
)
return [
DefaultInfo(
executable = outfile,
data_runfiles = actual_binary[DefaultInfo].data_runfiles,
),
]
# The purpose of this rule is to take a "set_flags" attribute, invoke a transition that sets
# any of _flags to the specified values, then depend on a cc_binary whose deps will be able
# to select() on those flags as if the user had set them via the CLI.
transition_rule = rule(
implementation = _transition_rule_impl,
attrs = {
# set_flags is a dictionary with the keys being the short-form of a flag name
# (e.g. the part that comes after the colon) and the value being a list of values
# that the flag should be set to, regardless of the relevant CLI flags.
# https://docs.bazel.build/versions/main/skylark/lib/attr.html#string_list_dict
"set_flags": attr.string_list_dict(),
# This is the cc_binary whose deps will select() on that feature.
# Note specifically how it is modified with _flag_transition, which
# ensures that the flags propagates down the graph.
# https://docs.bazel.build/versions/main/skylark/lib/attr.html#label
"actual_binary": attr.label(cfg = _flag_transition),
# This is a stock Bazel requirement for any rule that uses Starlark
# transitions. It's okay to copy the below verbatim for all such rules.
#
# The purpose of this requirement is to give the ability to restrict
# which packages can invoke these rules, since Starlark transitions
# make much larger graphs possible that can have memory and performance
# consequences for your build. The allowlist defaults to "everything".
# But you can redefine it more strictly if you feel that's prudent.
"_allowlist_function_transition": attr.label(
default = "@bazel_tools//tools/allowlists/function_transition_allowlist",
),
},
# Making this executable means it works with "$ bazel run".
executable = True,
)
def cc_binary_with_flags(name, cc_binary_name = "", set_flags = {}, **kwargs):
"""Builds a cc_binary as if set_flags were set on the CLI.
Args:
name: string, the name for the rule that is the binary, but with the flags changed via
a transition. Any dependents should use this name.
cc_binary_name: string, the name of the binary created (not the name of the transition
rule). The default is the name with "_native_binary" as a suffix.
set_flags: dictionary of string to list of strings. The keys should be the name of the
flag, and the values should be the desired valid settings for that flag.
**kwargs: Any flags that a cc_binary normally takes.
"""
if not cc_binary_name:
cc_binary_name = name + "_native_binary"
transition_rule(
name = name,
actual_binary = ":%s" % cc_binary_name,
set_flags = set_flags,
testonly = kwargs.get("testonly", False),
)
native.cc_binary(
name = cc_binary_name,
**kwargs
)