| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
TL;DR: The binding generator is a mess as of right now. At first it was funny
(in a "this is challenging" sense) to improve on it, but this is not
sustainable.
The truth is that the current architecture of the binding generator is a huge
pile of hacks, so these few days I've been working on rewriting it with a few
goals.
1) Have the hacks as contained and identified as possible. They're sometimes
needed because how clang exposes the AST, but ideally those hacks are well
identified and don't interact randomly with each others.
As an example, in the current bindgen when scanning the parameters of a
function that references a struct clones all the struct information, then if
the struct name changes (because we mangle it), everything breaks.
2) Support extending the bindgen output without having to deal with clang. The
way I'm aiming to do this is separating completely the parsing stage from
the code generation one, and providing a single id for each item the binding
generator provides.
3) No more random mutation of the internal representation from anywhere. That
means no more Rc<RefCell<T>>, no more random circular references, no more
borrow_state... nothing.
4) No more deduplication of declarations before code generation.
Current bindgen has a stage, called `tag_dup_decl`[1], that takes care of
deduplicating declarations. That's completely buggy, and for C++ it's a
complete mess, since we YOLO modify the world.
I've managed to take rid of this using the clang canonical declaration, and
the definition, to avoid scanning any type/item twice.
5) Code generation should not modify any internal data structure. It can lookup
things, traverse whatever it needs, but not modifying randomly.
6) Each item should have a canonical name, and a single source of mangling
logic, and that should be computed from the inmutable state, at code
generation.
I've put a few canonical_name stuff in the code generation phase, but it's
still not complete, and should change if I implement namespaces.
Improvements pending until this can land:
1) Add support for missing core stuff, mainly generating functions (note that
we parse the signatures for types correctly though), bitfields, generating
C++ methods.
2) Add support for the necessary features that were added to work around some
C++ pitfalls, like opaque types, etc...
3) Add support for the sugar that Manish added recently.
4) Optionally (and I guess this can land without it, because basically nobody
uses it since it's so buggy), bring back namespace support.
These are not completely trivial, but I think I can do them quite easily with
the current architecture.
I'm putting the current state of affairs here as a request for comments... Any
thoughts? Note that there are still a few smells I want to eventually
re-redesign, like the ParseError::Recurse thing, but until that happens I'm
way happier with this kind of architecture.
I'm keeping the old `parser.rs` and `gen.rs` in tree just for reference while I
code, but they will go away.
[1]: https://github.com/Yamakaky/rust-bindgen/blob/master/src/gen.rs#L448
|
|
|
|
There's no easy way to get the current target triple from the clang-c
bindings. Make this simple and require an option to be passed when
generating bindings for msvc, using the x86_64-pc-win32 target (or i686)
|
|
This is not needed at least for Clang 3.9 (which is required to parse
stdlib headers of MSVC 2015). This hack leads to incorrect mangling
result there.
|
|
Fixes #1.
|
|
|
|
Ignore inlined functions
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Generate better enums
Squash of...
Disable prefixing
Default to failing on unknown types
Add support for Char16
Emit errors for unknown cursor kinds
Hack in support for classes
Recurse into unexposed decls
This fixes functions that use extern "C".
Add support for generating unmangled functions
Prefix unnamed data structures with the file name
Recurse into namespaced things
Avoid rust reserved keywords in unmangle func args
Don't create variadic unmangled funcs
Don't translate typedefs to the same name
Ignore operator overloads
Avoid templates
Handle class declarations
Number duplicate demangle functions
Implement copy on enums
Translate stdint types into standard rust int types
Switch enums to i32 for better compatibility
Correctly deal with mangled functions with unnamed args
Mark unmangling functions as unsafe
Attempt to produce structs for C++ classes
Convert references
Generate better func decls for void returns
Make every function callback unsafe
Add support for generics in typedefs
Add support for class templates
Aggressively trim duplicates
Don't generate default impl for templates
Improve handling of templates
Fix function numbering
Fix deduplication
Make unmangling functions extern "C"
Convert all int/float typedefs to standard rust ints/floats
This also gives better information to the bitfield parsing and allows uint32_t and other stdint bitfields to be processed properly
Add support for wchar
Add support for bitfield setter generation
Fix floats
Squash of...
Shorten generated bitfield names
Add support for generating whole bitfields
Add support for enums nested inside structs/classes
Rustup
Fixes #184.
Rustup to b301e02f3 2015-05-19
Inline demangling functions
Add support for base classes/types
Generate bindings for methods
Make duplicate detection less aggressive
Avoid converting long/unsigned longs to rust types.
This fixes 64/32bit issues in structs.
Generate bitfields correctly for typedefs
Convert stdint types to rust types
Derive Debug on BindgenOptions, Bindings, and LinkType.
Remove +'static when writing bindings
Generate virtual function tables
Resolve some warnings
Add NotConst where Constness params are required
Generate real bools when applicable
Squash of...
Add support for comments
Improve bitfield support using const fn
Add limited support for references
Add comments to fields
Don't generate empty comments
Convert char16_t to u16 rather than i16
Convert signed chars to libc::c_char
Fix Cargo.toml rebasing breakage
Fix compiler errors
This gets bindgen to compile and run again, but all but one `cargo
test` tests fail. Not sure if that’s because of mistakes on my part or
if the sm-hacks branch never passed those tests.
Fix build warnings
Use link_name attr for mangled function names
Handle mangled global vars
Only generate bindings for visible symbols
Don't generate arrays for blobs when the length is 1
Name enums inside classes better
Handle template args inside typedefs better
Filter out duplicate decls better
Generate correctly sized enums
Squash of...
Fix bitfield accessor method generation for bools
Insert phantom fields in empty structs
Don't mark unmangling methods as extern "C"
Add back comment support for functions
Add basic annotation support
Don't generate univariant enums
Add support for hide annotation and adjust syntax
Don't generate unsupported structs
Don't parse hidden fields
Don't derive Copy for structs with destructors
Don't implement Clone or Default
Derive Clone when deriving Copy
Bypass single member unions
Disable references in function args for now
Remove extra underscore in mangled names on OSX
Don't translate private methods
Support generating consts from macros that are defined as integer literals.
Handle mangling better
Squash of...
Update README.md for fork
Generate docs for enum items
Generate docs for typedefs
Generate docs for enums
Update syntex_syntax to 0.24.*
Update clang info in README.md
Spit errors and warnings to stdout.
The correct thing to do here is to use env_logger, but that was causing cargo
troubles for me, and this is preferable to swallowing them.
Add the -ignore-functions argument.
Handle 1 << 63 as enum value.
Don't try to convert standard int types in rust_type_id.
It looks like mwu added this, but I'm pretty sure it's a category error. This
function appears to be designed to reproducibly permute C identifiers so that
they don't conflict with builtin rust types. It's specifically _not_ a type
translator (which would happen at the type level, rather than the string
level), and using it as such with callers like ctypedef_to_rs causes us to
generate things like:
type u64 = u64;
While processing stdint.h, which is clearly wrong.
Stop patching in placeholder names for CompInfo and EnumInfo instances during code generator.
As best as I can tell, it's done this way (rather than my way) because bindgen tries
to recognize struct and enums typedefs of the form:
/* This is a common idiom in C, not so much in C++ */
typdef struct {
...
} Foo;
The intention, I think, is to avoid generating rust code for a struct with a placeholder
name followed by a typedef, and just give the struct the right name initially.
This seems like a reasonable goal, though not a particularly important one. However, in
my testing this never actually happens, because we end up calling unnamed_name anyway
during the GComp(ci) case of gen_mod before we get to evaluting the typedef.
So let's just remove that stuff and simplify the code. This lets us remove all the
borrow_mut calls during code generation, which seems necessary for soundness.
gen: Allow empty union members
Use full paths in generation.
Fix test compilation
parser: Add support for parsing namespaces
Partial C++ namespaces support
We currently generate the modules as expected, but we don't resolve the
names from external namespaces well.
Remove unnecesary return statements
Put namespaces behind a flag
Overall now that they aren't complete still.
Moar refactoring
Finally take rid of all the warnings
Even moar
gen: Avoid so much cloning
parser: Refactor the way submodules are stored
This way we can share the global map, while having each module custom
globals.
gen: Checkpoint before the refactoring
This actually keeps working as before.
gen: Make modules (almost) work for typedef'd types
We generate almost valid code, we just have to add some use statements.
Or maybe is a better idea to add an unintelligible name to the root
mod, and actually output a root mod plus a use root::* before.
gen: Document the current submodule approach and some desirable
alternative
gen: Make it actually compilable \o/
gen: Make the code generation usable for every type.
There's just an edge case I've detected, and it's when we remove the
instantiation of C<int>, and another module uses it, because that means
we only know of its existance in that other module.
Probably we might want to use cursor_getSemanticParent to get the real
class instead of the instantiated, but I'm not too confident about that.
Fix a corner case when a template was instantiated in another module.
Added an example of the namespace resolution.
Don't panic when not finding the specialised template
This can be annoying if filtering files out.
Straight rebase completed, let's fix that build errors
wip
Pair up with master
nits
Update AST
Add -no-rename-fields option
This is for compatibility between C bindings and C++ bindings (in C
`struct Foo` and `enum Foo`, are different, while in C++ they aren't).
wip
Add enum tests pass, and add C++ tests
Make a few more struct-related tests pass
|
|
|
|
|
|
|
|
|
|
This patch adds a few things to enable this:
* Cursor::extent
* struct Token
* TranslationUnit::tokens
- clang's C API doesn't have a straightforward way to get the integer
literal from a variable declaration, so I used clang_tokenize to
get the literal token and parse it.
* VarInfo::val - to store the value.
Non-const variable declarations and non-integer constants are handled as
they were previously.
|
|
|
|
|
|
|
|
Fixes #184.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
clang.rs
- Remove Box<_> from the signature of Cursor.visit. This makes the
implementation a little noisier, but makes calling easier.
tests
- Updated for recent changes to rustc and std.
|
|
|
|
|
|
|
|
|
|
can be written by hand if required.
|
|
|
|
|
|
|
|
With this patch, rust-bindgen generates extern functions with
calling conventions specified in header.
This patch also removes `-abi` flag.
|
|
|
|
|
|
|
