FFI baked data

ffi/capi_cdylib/Cargo.toml

@@ -39,6 +39,7 @@ icu_provider = { version = "1.0.0", path = "../../provider/core", default-featur
 # Please keep features/cargo-all-features lists in sync with the icu_capi crate
 [features]
 default = ["icu_capi/default"]
+custom_bake_data = ["icu_capi/custom_bake_data"]
 any_provider = ["icu_capi/any_provider"]
 buffer_provider = ["icu_capi/buffer_provider"]
 provider_fs = ["icu_capi/provider_fs"] # Indirectly implies buffer_provider

ffi/capi_staticlib/Cargo.toml

@@ -39,6 +39,7 @@ icu_provider = { version = "1.0.0", path = "../../provider/core", default-featur
 # Please keep features/cargo-all-features lists in sync with the icu_capi crate
 [features]
 default = ["icu_capi/default"]
+custom_bake_data = ["icu_capi/custom_bake_data"]
 any_provider = ["icu_capi/any_provider"]
 buffer_provider = ["icu_capi/buffer_provider"]
 provider_fs = ["icu_capi/provider_fs"] # Indirectly implies buffer_provider

ffi/diplomat/Cargo.toml

@@ -30,11 +30,12 @@ all-features = true
 [package.metadata.cargo-all-features]
 # Bench feature gets tested separately and is only relevant for CI.
 # logging enables a feature of a dependency that has no externally visible API changes
-denylist = ["bench", "logging", "cpp_default", "wasm_default"]
+denylist = ["bench", "logging", "cpp_default", "wasm_default", "custom_bake_data"]
 
 # Please keep the features list in sync with the icu_capi_staticlib/icu_capi_cdylib crates
 [features]
-default = ["any_provider"]
+default = ["custom_bake_data"]
+custom_bake_data = []
 any_provider = []
 buffer_provider = [
     "dep:icu_provider_blob",
@@ -60,7 +61,7 @@ logging = ["icu_provider/log_error_context", "dep:log"]
 simple_logger = ["dep:simple_logger"]
 
 # meta feature for things we enable by default in C and C++
-cpp_default = ["provider_test", "logging", "simple_logger"]
+cpp_default = ["provider_test", "any_provider", "logging", "simple_logger"]
 
 # meta feature for things we enable by default in wasm
 wasm_default = ["buffer_provider", "logging"]
@@ -91,6 +92,9 @@ serde = { version = "1.0", default-features = false, optional = true }
 
 icu_testdata = { version = "1.0.0", path = "../../provider/testdata", optional = true, features = ["icu_segmenter"] }
 
+# databake
+zerovec = { version = "*", path = "../../utils/zerovec" }
+
 # Run `cargo make diplomat-install` to get the right diplomat binary installed
 # The version here can either be a `version = ".."` spec or `git = "https://github.com/rust-diplomat/diplomat", rev = ".."`
 # Since this crate is published, Diplomat must be published preceding a new ICU4X release but may use git versions in between

ffi/diplomat/build.rs

@@ -0,0 +1,11 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+fn main() {
+    if std::env::var("ICU4X_FFI_BAKED_ROOT").is_err() {
+        // Empty data generated with
+        // cargo run -p icu_datagen -- --format mod --use-separate-crates --keys none --locales none --out empty_bake
+        println!("cargo:rustc-env=ICU4X_FFI_BAKED_ROOT=../empty_bake");
+    }
+}

ffi/diplomat/c/examples/fixeddecimal_tiny/.gitignore

@@ -5,5 +5,4 @@ optim*
 *.elf
 *.o
 a.out.dSYM
-decimal-bn-en.postcard
-decimal_bn_en.h
+baked

ffi/diplomat/c/examples/fixeddecimal_tiny/Makefile

@@ -17,51 +17,49 @@ GCC := gcc
 CLANG := clang-14
 LLD := lld-14
 
+export ICU4X_FFI_BAKED_ROOT = $(dir $(realpath $(lastword $(MAKEFILE_LIST))))baked
 
-../../../../../target/debug/libicu_capi_staticlib.a: $(ALL_RUST)
-	cargo build -p icu_capi_staticlib --no-default-features --features buffer_provider
+../../../../../target/debug/libicu_capi_staticlib.a: $(ALL_RUST) $(ICU4X_FFI_BAKED_ROOT)
+	cargo build -p icu_capi_staticlib
 
-icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a: $(ALL_RUST)
+icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a: $(ALL_RUST) $(ICU4X_FFI_BAKED_ROOT)
 	cd icu_capi_staticlib_tiny && \
-	RUSTFLAGS="-Clinker-plugin-lto -Clinker=$(CLANG) -Ccodegen-units=1 -Clink-arg=-flto -Cpanic=abort" cargo +${ICU4X_NIGHTLY_TOOLCHAIN} panic-abort-build --target x86_64-unknown-linux-gnu
+	RUSTFLAGS="-Clinker-plugin-lto -Clinker=$(CLANG) -Ccodegen-units=1 -Clink-arg=-flto -Cpanic=abort" cargo +$(ICU4X_NIGHTLY_TOOLCHAIN) panic-abort-build --target x86_64-unknown-linux-gnu
 
-icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/release-opt-size/libicu_capi_staticlib_tiny.a: $(ALL_RUST)
+icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/release-opt-size/libicu_capi_staticlib_tiny.a: $(ALL_RUST) $(ICU4X_FFI_BAKED_ROOT)
 	cd icu_capi_staticlib_tiny && \
-	RUSTFLAGS="-Clinker-plugin-lto -Clinker=$(CLANG) -Ccodegen-units=1 -Clink-arg=-flto -Cpanic=abort" cargo +${ICU4X_NIGHTLY_TOOLCHAIN} panic-abort-build --target x86_64-unknown-linux-gnu --profile=release-opt-size
+	RUSTFLAGS="-Clinker-plugin-lto -Clinker=$(CLANG) -Ccodegen-units=1 -Clink-arg=-flto -Cpanic=abort" cargo +$(ICU4X_NIGHTLY_TOOLCHAIN) panic-abort-build --target x86_64-unknown-linux-gnu --profile=release-opt-size
 
-decimal-bn-en.postcard:
-	cargo run -p icu_datagen -- --locales en bn --keys "decimal/symbols@1" --cldr-root ../../../../../provider/testdata/data/cldr/ --format blob --out decimal-bn-en.postcard
-
-decimal_bn_en.h: decimal-bn-en.postcard
-	xxd -i -C decimal-bn-en.postcard > decimal_bn_en.h
+$(ICU4X_FFI_BAKED_ROOT):
+	cargo run -p icu_datagen -- --locales en bn --keys "decimal/symbols@1" --cldr-root ../../../../../provider/testdata/data/cldr/ --format mod --use-separate-crates --out ${ICU4X_FFI_BAKED_ROOT}
 
 # Naive target: no optimizations, full std
-optim0.elf: ../../../../../target/debug/libicu_capi_staticlib.a $(ALL_HEADERS) test.c decimal_bn_en.h
+optim0.elf: ../../../../../target/debug/libicu_capi_staticlib.a $(ALL_HEADERS) test.c
 	$(GCC) test.c ../../../../../target/debug/libicu_capi_staticlib.a -ldl -lpthread -lm -g -o optim0.elf
 
 # optim.elf: gcc with maximum link-time code stripping (gc-sections and strip-all)
-optim1.elf: ../../../../../target/debug/libicu_capi_staticlib.a $(ALL_HEADERS) test.c decimal_bn_en.h
+optim1.elf: ../../../../../target/debug/libicu_capi_staticlib.a $(ALL_HEADERS) test.c
 	$(GCC) -fdata-sections -ffunction-sections test.c ../../../../../target/debug/libicu_capi_staticlib.a -ldl -lpthread -lm -g -o optim1.elf -Wl,--gc-sections -Wl,--strip-all
 
 # optim2.elf: clang single-step with gc-sections
-optim2.elf: icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a $(ALL_HEADERS) test.c decimal_bn_en.h
+optim2.elf: icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a $(ALL_HEADERS) test.c
 	$(CLANG) -flto -fdata-sections -ffunction-sections test.c icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a -g -o optim2.elf -Wl,--gc-sections
 
-optim3.o: $(ALL_HEADERS) test.c decimal_bn_en.h
+optim3.o: $(ALL_HEADERS) test.c
 	$(CLANG) -c -flto=thin -fdata-sections -ffunction-sections --target=x86_64-unknown-linux-gnu test.c -g -o optim3.o
 
 # optim3.elf: clang two-step with lld, debug mode
 optim3.elf: optim3.o icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a
 	$(CLANG) -flto=thin -fuse-ld=$(LLD) -L . -o optim3.elf optim3.o icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/debug/libicu_capi_staticlib_tiny.a -Wl,--gc-sections
 
-optim4.o: $(ALL_HEADERS) test.c decimal_bn_en.h
+optim4.o: $(ALL_HEADERS) test.c
 	$(CLANG) -c -flto=thin -fdata-sections -ffunction-sections --target=x86_64-unknown-linux-gnu test.c -g -o optim4.o
 
 # optim4.elf: clang two-step with lld, release mode with debug symbols
 optim4.elf: optim4.o icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/release-opt-size/libicu_capi_staticlib_tiny.a
 	$(CLANG) -flto=thin -fuse-ld=$(LLD) -L . -o optim4.elf optim4.o icu_capi_staticlib_tiny/target/x86_64-unknown-linux-gnu/release-opt-size/libicu_capi_staticlib_tiny.a -Wl,--gc-sections
 
-optim5.o: $(ALL_HEADERS) test.c decimal_bn_en.h
+optim5.o: $(ALL_HEADERS) test.c
 	$(CLANG) -c -flto=thin -fdata-sections -ffunction-sections --target=x86_64-unknown-linux-gnu test.c -o optim5.o
 
 # optim5.elf: clang two-step with lld, release mode stripped of debug symbols