diff options
author | DongHun Kwak <dh0128.kwak@samsung.com> | 2023-05-17 16:16:15 +0900 |
---|---|---|
committer | DongHun Kwak <dh0128.kwak@samsung.com> | 2023-05-17 16:16:15 +0900 |
commit | d152277e566567fe8d6ed10aec3ed1b52a0b9407 (patch) | |
tree | dcc129fa9a17dd92887f21ed8a5f1f1ec132f826 | |
download | rust-bytecheck_derive-d152277e566567fe8d6ed10aec3ed1b52a0b9407.tar.gz rust-bytecheck_derive-d152277e566567fe8d6ed10aec3ed1b52a0b9407.tar.bz2 rust-bytecheck_derive-d152277e566567fe8d6ed10aec3ed1b52a0b9407.zip |
Import bytecheck_derive 0.7.0upstream/0.7.0upstream
-rw-r--r-- | .cargo_vcs_info.json | 6 | ||||
-rw-r--r-- | Cargo.toml | 37 | ||||
-rw-r--r-- | Cargo.toml.orig | 23 | ||||
-rw-r--r-- | LICENSE | 7 | ||||
-rw-r--r-- | src/lib.rs | 533 |
5 files changed, 606 insertions, 0 deletions
diff --git a/.cargo_vcs_info.json b/.cargo_vcs_info.json new file mode 100644 index 0000000..49aa54f --- /dev/null +++ b/.cargo_vcs_info.json @@ -0,0 +1,6 @@ +{ + "git": { + "sha1": "e37fd31f50b2fd7778601999922534ba43c36b14" + }, + "path_in_vcs": "bytecheck_derive" +}
\ No newline at end of file diff --git a/Cargo.toml b/Cargo.toml new file mode 100644 index 0000000..96569e5 --- /dev/null +++ b/Cargo.toml @@ -0,0 +1,37 @@ +# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO +# +# When uploading crates to the registry Cargo will automatically +# "normalize" Cargo.toml files for maximal compatibility +# with all versions of Cargo and also rewrite `path` dependencies +# to registry (e.g., crates.io) dependencies. +# +# If you are reading this file be aware that the original Cargo.toml +# will likely look very different (and much more reasonable). +# See Cargo.toml.orig for the original contents. + +[package] +edition = "2021" +name = "bytecheck_derive" +version = "0.7.0" +authors = ["David Koloski <djkoloski@gmail.com>"] +description = "Derive macro for bytecheck" +documentation = "https://docs.rs/bytecheck_derive" +license = "MIT" +repository = "https://github.com/djkoloski/bytecheck" +resolver = "1" + +[lib] +proc-macro = true + +[dependencies.proc-macro2] +version = "1.0" + +[dependencies.quote] +version = "1.0" + +[dependencies.syn] +version = "1.0" + +[features] +default = ["std"] +std = [] diff --git a/Cargo.toml.orig b/Cargo.toml.orig new file mode 100644 index 0000000..5a541c8 --- /dev/null +++ b/Cargo.toml.orig @@ -0,0 +1,23 @@ +[package]
+name = "bytecheck_derive"
+description = "Derive macro for bytecheck"
+documentation = "https://docs.rs/bytecheck_derive"
+version.workspace = true
+edition.workspace = true
+authors.workspace = true
+license.workspace = true
+repository.workspace = true
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[lib]
+proc-macro = true
+
+[dependencies]
+proc-macro2 = "1.0"
+syn = "1.0"
+quote = "1.0"
+
+[features]
+default = ["std"]
+std = []
@@ -0,0 +1,7 @@ +Copyright 2020 David Koloski
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
diff --git a/src/lib.rs b/src/lib.rs new file mode 100644 index 0000000..ef61982 --- /dev/null +++ b/src/lib.rs @@ -0,0 +1,533 @@ +//! Procedural macros for bytecheck.
+
+#![deny(
+ rust_2018_compatibility,
+ rust_2018_idioms,
+ future_incompatible,
+ nonstandard_style,
+ unused,
+ clippy::all
+)]
+
+use proc_macro2::{Group, Span, TokenStream, TokenTree};
+use quote::{quote, quote_spanned};
+use syn::{
+ parse_macro_input, parse_quote, punctuated::Punctuated, spanned::Spanned, AttrStyle, Data,
+ DeriveInput, Error, Fields, Ident, Index, Lit, LitStr, Meta, NestedMeta, Path, Token,
+ WherePredicate,
+};
+
+#[derive(Default)]
+struct Repr {
+ pub transparent: Option<Path>,
+ pub packed: Option<Path>,
+ pub c: Option<Path>,
+ pub int: Option<Path>,
+}
+
+#[derive(Default)]
+struct Attributes {
+ pub repr: Repr,
+ pub bound: Option<LitStr>,
+ pub bytecheck_crate: Option<Path>,
+}
+
+fn parse_check_bytes_attributes(attributes: &mut Attributes, meta: &Meta) -> Result<(), Error> {
+ match meta {
+ Meta::NameValue(meta) => {
+ if meta.path.is_ident("bound") {
+ if let Lit::Str(ref lit_str) = meta.lit {
+ if attributes.bound.is_none() {
+ attributes.bound = Some(lit_str.clone());
+ Ok(())
+ } else {
+ Err(Error::new_spanned(
+ meta,
+ "check_bytes bound already specified",
+ ))
+ }
+ } else {
+ Err(Error::new_spanned(
+ &meta.lit,
+ "bound arguments must be a string",
+ ))
+ }
+ } else if meta.path.is_ident("crate") {
+ if let Lit::Str(ref lit_str) = meta.lit {
+ if attributes.bytecheck_crate.is_none() {
+ let tokens = respan(syn::parse_str(&lit_str.value())?, lit_str.span());
+ let parsed: Path = syn::parse2(tokens)?;
+ attributes.bytecheck_crate = Some(parsed);
+ Ok(())
+ } else {
+ Err(Error::new_spanned(
+ meta,
+ "check_bytes crate already specified",
+ ))
+ }
+ } else {
+ Err(Error::new_spanned(
+ &meta.lit,
+ "crate argument must be a string",
+ ))
+ }
+ } else {
+ Err(Error::new_spanned(
+ &meta.path,
+ "unrecognized check_bytes argument",
+ ))
+ }
+ }
+ _ => Err(Error::new_spanned(
+ meta,
+ "unrecognized check_bytes argument",
+ )),
+ }
+}
+
+fn parse_attributes(input: &DeriveInput) -> Result<Attributes, Error> {
+ let mut result = Attributes::default();
+ for a in input.attrs.iter() {
+ if let AttrStyle::Outer = a.style {
+ if let Ok(Meta::List(meta)) = a.parse_meta() {
+ if meta.path.is_ident("check_bytes") {
+ for nested in meta.nested.iter() {
+ if let NestedMeta::Meta(meta) = nested {
+ parse_check_bytes_attributes(&mut result, meta)?;
+ } else {
+ return Err(Error::new_spanned(
+ nested,
+ "check_bytes parameters must be metas",
+ ));
+ }
+ }
+ } else if meta.path.is_ident("repr") {
+ for n in meta.nested.iter() {
+ if let NestedMeta::Meta(Meta::Path(path)) = n {
+ if path.is_ident("transparent") {
+ result.repr.transparent = Some(path.clone());
+ } else if path.is_ident("packed") {
+ result.repr.packed = Some(path.clone());
+ } else if path.is_ident("C") {
+ result.repr.c = Some(path.clone());
+ } else if path.is_ident("align") {
+ // Ignore alignment modifiers
+ } else {
+ let is_int_repr = path.is_ident("i8")
+ || path.is_ident("i16")
+ || path.is_ident("i32")
+ || path.is_ident("i64")
+ || path.is_ident("i128")
+ || path.is_ident("u8")
+ || path.is_ident("u16")
+ || path.is_ident("u32")
+ || path.is_ident("u64")
+ || path.is_ident("u128");
+
+ if is_int_repr {
+ result.repr.int = Some(path.clone());
+ } else {
+ return Err(Error::new_spanned(
+ path,
+ "invalid repr, available reprs are transparent, C, i* and u*",
+ ));
+ }
+ }
+ }
+ }
+ }
+ }
+ }
+ }
+ Ok(result)
+}
+
+/// Derives `CheckBytes` for the labeled type.
+///
+/// Additional arguments can be specified using the `#[check_bytes(...)]` attribute:
+///
+/// - `bound = "..."`: Adds additional bounds to the `CheckBytes` implementation. This can be
+/// especially useful when dealing with recursive structures, where bounds may need to be omitted
+/// to prevent recursive type definitions.
+///
+/// This derive macro automatically adds a type bound `field: CheckBytes<__C>` for each field type.
+/// This can cause an overflow while evaluating trait bounds if the structure eventually references
+/// its own type, as the implementation of `CheckBytes` for a struct depends on each field type
+/// implementing it as well. Adding the attribute `#[omit_bounds]` to a field will suppress this
+/// trait bound and allow recursive structures. This may be too coarse for some types, in which case
+/// additional type bounds may be required with `bound = "..."`.
+#[proc_macro_derive(CheckBytes, attributes(check_bytes, omit_bounds))]
+pub fn check_bytes_derive(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
+ match derive_check_bytes(parse_macro_input!(input as DeriveInput)) {
+ Ok(result) => result.into(),
+ Err(e) => e.to_compile_error().into(),
+ }
+}
+
+fn derive_check_bytes(mut input: DeriveInput) -> Result<TokenStream, Error> {
+ let attributes = parse_attributes(&input)?;
+
+ let mut impl_input_generics = input.generics.clone();
+ let impl_where_clause = impl_input_generics.make_where_clause();
+ if let Some(ref bounds) = attributes.bound {
+ let clauses =
+ bounds.parse_with(Punctuated::<WherePredicate, Token![,]>::parse_terminated)?;
+ for clause in clauses {
+ impl_where_clause.predicates.push(clause);
+ }
+ }
+ impl_input_generics
+ .params
+ .insert(0, parse_quote! { __C: ?Sized });
+
+ let name = &input.ident;
+
+ let (impl_generics, _, impl_where_clause) = impl_input_generics.split_for_impl();
+ let impl_where_clause = impl_where_clause.unwrap();
+
+ input.generics.make_where_clause();
+ let (struct_generics, ty_generics, where_clause) = input.generics.split_for_impl();
+ let where_clause = where_clause.unwrap();
+
+ let check_bytes_impl = match input.data {
+ Data::Struct(ref data) => match data.fields {
+ Fields::Named(ref fields) => {
+ let mut check_where = impl_where_clause.clone();
+ for field in fields
+ .named
+ .iter()
+ .filter(|f| !f.attrs.iter().any(|a| a.path.is_ident("omit_bounds")))
+ {
+ let ty = &field.ty;
+ check_where
+ .predicates
+ .push(parse_quote! { #ty: CheckBytes<__C> });
+ }
+
+ let field_checks = fields.named.iter().map(|f| {
+ let field = &f.ident;
+ let ty = &f.ty;
+ quote_spanned! { ty.span() =>
+ <#ty as CheckBytes<__C>>::check_bytes(
+ ::core::ptr::addr_of!((*value).#field),
+ context
+ ).map_err(|e| StructCheckError {
+ field_name: stringify!(#field),
+ inner: ErrorBox::new(e),
+ })?;
+ }
+ });
+
+ quote! {
+ #[automatically_derived]
+ impl #impl_generics CheckBytes<__C> for #name #ty_generics #check_where {
+ type Error = StructCheckError;
+
+ unsafe fn check_bytes<'__bytecheck>(
+ value: *const Self,
+ context: &mut __C,
+ ) -> ::core::result::Result<&'__bytecheck Self, StructCheckError> {
+ let bytes = value.cast::<u8>();
+ #(#field_checks)*
+ Ok(&*value)
+ }
+ }
+ }
+ }
+ Fields::Unnamed(ref fields) => {
+ let mut check_where = impl_where_clause.clone();
+ for field in fields
+ .unnamed
+ .iter()
+ .filter(|f| !f.attrs.iter().any(|a| a.path.is_ident("omit_bounds")))
+ {
+ let ty = &field.ty;
+ check_where
+ .predicates
+ .push(parse_quote! { #ty: CheckBytes<__C> });
+ }
+
+ let field_checks = fields.unnamed.iter().enumerate().map(|(i, f)| {
+ let ty = &f.ty;
+ let index = Index::from(i);
+ quote_spanned! { ty.span() =>
+ <#ty as CheckBytes<__C>>::check_bytes(
+ ::core::ptr::addr_of!((*value).#index),
+ context
+ ).map_err(|e| TupleStructCheckError {
+ field_index: #i,
+ inner: ErrorBox::new(e),
+ })?;
+ }
+ });
+
+ quote! {
+ #[automatically_derived]
+ impl #impl_generics CheckBytes<__C> for #name #ty_generics #check_where {
+ type Error = TupleStructCheckError;
+
+ unsafe fn check_bytes<'__bytecheck>(
+ value: *const Self,
+ context: &mut __C,
+ ) -> ::core::result::Result<&'__bytecheck Self, TupleStructCheckError> {
+ let bytes = value.cast::<u8>();
+ #(#field_checks)*
+ Ok(&*value)
+ }
+ }
+ }
+ }
+ Fields::Unit => {
+ quote! {
+ #[automatically_derived]
+ impl #impl_generics CheckBytes<__C> for #name #ty_generics #impl_where_clause {
+ type Error = Infallible;
+
+ unsafe fn check_bytes<'__bytecheck>(
+ value: *const Self,
+ context: &mut __C,
+ ) -> ::core::result::Result<&'__bytecheck Self, Infallible> {
+ Ok(&*value)
+ }
+ }
+ }
+ }
+ },
+ Data::Enum(ref data) => {
+ if let Some(path) = attributes.repr.transparent.or(attributes.repr.packed) {
+ return Err(Error::new_spanned(
+ path,
+ "enums implementing CheckBytes cannot be repr(transparent) or repr(packed)",
+ ));
+ }
+
+ let repr = match attributes.repr.int {
+ None => {
+ return Err(Error::new(
+ input.span(),
+ "enums implementing CheckBytes must be repr(Int)",
+ ));
+ }
+ Some(ref repr) => repr,
+ };
+
+ let mut check_where = impl_where_clause.clone();
+ for v in data.variants.iter() {
+ match v.fields {
+ Fields::Named(ref fields) => {
+ for field in fields
+ .named
+ .iter()
+ .filter(|f| !f.attrs.iter().any(|a| a.path.is_ident("omit_bounds")))
+ {
+ let ty = &field.ty;
+ check_where
+ .predicates
+ .push(parse_quote! { #ty: CheckBytes<__C> });
+ }
+ }
+ Fields::Unnamed(ref fields) => {
+ for field in fields
+ .unnamed
+ .iter()
+ .filter(|f| !f.attrs.iter().any(|a| a.path.is_ident("omit_bounds")))
+ {
+ let ty = &field.ty;
+ check_where
+ .predicates
+ .push(parse_quote! { #ty: CheckBytes<__C> });
+ }
+ }
+ Fields::Unit => (),
+ }
+ }
+
+ let tag_variant_defs = data.variants.iter().map(|v| {
+ let variant = &v.ident;
+ if let Some((_, expr)) = &v.discriminant {
+ quote_spanned! { variant.span() => #variant = #expr }
+ } else {
+ quote_spanned! { variant.span() => #variant }
+ }
+ });
+
+ let discriminant_const_defs = data.variants.iter().map(|v| {
+ let variant = &v.ident;
+ quote! {
+ #[allow(non_upper_case_globals)]
+ const #variant: #repr = Tag::#variant as #repr;
+ }
+ });
+
+ let tag_variant_values = data.variants.iter().map(|v| {
+ let name = &v.ident;
+ quote_spanned! { name.span() => Discriminant::#name }
+ });
+
+ let variant_structs = data.variants.iter().map(|v| {
+ let variant = &v.ident;
+ let variant_name = Ident::new(&format!("Variant{variant}"), v.span());
+ match v.fields {
+ Fields::Named(ref fields) => {
+ let fields = fields.named.iter().map(|f| {
+ let name = &f.ident;
+ let ty = &f.ty;
+ quote_spanned! { f.span() => #name: #ty }
+ });
+ quote_spanned! { name.span() =>
+ #[repr(C)]
+ struct #variant_name #struct_generics #where_clause {
+ __tag: Tag,
+ #(#fields,)*
+ __phantom: PhantomData<#name #ty_generics>,
+ }
+ }
+ }
+ Fields::Unnamed(ref fields) => {
+ let fields = fields.unnamed.iter().map(|f| {
+ let ty = &f.ty;
+ quote_spanned! { f.span() => #ty }
+ });
+ quote_spanned! { name.span() =>
+ #[repr(C)]
+ struct #variant_name #struct_generics (
+ Tag,
+ #(#fields,)*
+ PhantomData<#name #ty_generics>
+ ) #where_clause;
+ }
+ }
+ Fields::Unit => quote! {},
+ }
+ });
+
+ let check_arms = data.variants.iter().map(|v| {
+ let variant = &v.ident;
+ let variant_name = Ident::new(&format!("Variant{variant}"), v.span());
+ match v.fields {
+ Fields::Named(ref fields) => {
+ let checks = fields.named.iter().map(|f| {
+ let name = &f.ident;
+ let ty = &f.ty;
+ quote! {
+ <#ty as CheckBytes<__C>>::check_bytes(
+ ::core::ptr::addr_of!((*value).#name),
+ context
+ ).map_err(|e| EnumCheckError::InvalidStruct {
+ variant_name: stringify!(#variant),
+ inner: StructCheckError {
+ field_name: stringify!(#name),
+ inner: ErrorBox::new(e),
+ },
+ })?;
+ }
+ });
+ quote_spanned! { variant.span() => {
+ let value = value.cast::<#variant_name #ty_generics>();
+ #(#checks)*
+ } }
+ }
+ Fields::Unnamed(ref fields) => {
+ let checks = fields.unnamed.iter().enumerate().map(|(i, f)| {
+ let ty = &f.ty;
+ let index = Index::from(i + 1);
+ quote! {
+ <#ty as CheckBytes<__C>>::check_bytes(
+ ::core::ptr::addr_of!((*value).#index),
+ context
+ ).map_err(|e| EnumCheckError::InvalidTuple {
+ variant_name: stringify!(#variant),
+ inner: TupleStructCheckError {
+ field_index: #i,
+ inner: ErrorBox::new(e),
+ },
+ })?;
+ }
+ });
+ quote_spanned! { variant.span() => {
+ let value = value.cast::<#variant_name #ty_generics>();
+ #(#checks)*
+ } }
+ }
+ Fields::Unit => quote_spanned! { name.span() => (), },
+ }
+ });
+
+ quote! {
+ #[repr(#repr)]
+ enum Tag {
+ #(#tag_variant_defs,)*
+ }
+
+ struct Discriminant;
+
+ #[automatically_derived]
+ impl Discriminant {
+ #(#discriminant_const_defs)*
+ }
+
+ #(#variant_structs)*
+
+ #[automatically_derived]
+ impl #impl_generics CheckBytes<__C> for #name #ty_generics #check_where {
+ type Error = EnumCheckError<#repr>;
+
+ unsafe fn check_bytes<'__bytecheck>(
+ value: *const Self,
+ context: &mut __C,
+ ) -> ::core::result::Result<&'__bytecheck Self, EnumCheckError<#repr>> {
+ let tag = *value.cast::<#repr>();
+ match tag {
+ #(#tag_variant_values => #check_arms)*
+ _ => return Err(EnumCheckError::InvalidTag(tag)),
+ }
+ Ok(&*value)
+ }
+ }
+ }
+ }
+ Data::Union(_) => {
+ return Err(Error::new(
+ input.span(),
+ "CheckBytes cannot be derived for unions",
+ ));
+ }
+ };
+
+ // Default to `bytecheck`, rather than `::bytecheck`,
+ // to allow providing it from a reexport, e.g. `use rkyv::bytecheck;`.
+ let bytecheck_crate = attributes
+ .bytecheck_crate
+ .unwrap_or(parse_quote!(bytecheck));
+
+ Ok(quote! {
+ #[allow(unused_results)]
+ const _: () = {
+ use ::core::{convert::Infallible, marker::PhantomData};
+ use #bytecheck_crate::{
+ CheckBytes,
+ EnumCheckError,
+ ErrorBox,
+ StructCheckError,
+ TupleStructCheckError,
+ };
+
+ #check_bytes_impl
+ };
+ })
+}
+
+fn respan(stream: TokenStream, span: Span) -> TokenStream {
+ stream
+ .into_iter()
+ .map(|token| respan_token(token, span))
+ .collect()
+}
+
+fn respan_token(mut token: TokenTree, span: Span) -> TokenTree {
+ if let TokenTree::Group(g) = &mut token {
+ *g = Group::new(g.delimiter(), respan(g.stream(), span));
+ }
+ token.set_span(span);
+ token
+}
|