failed to produce bindings for RPM library #363

sa2ajj · 2016-12-27T16:54:05Z

I'm looking into creating Rust bindings for the RPM libraries.

First I installed bindgen, but it seems to be somewhat inconsistent with what servo/rust-bindgen offers. So my second attempt was to clone and build bindgen from servo/rust-bindgen:

$ git clone https://github.com/servo/rust-bindgen
$ cd rust-bindgen/bindgen
$ cargo build --release

after that I created a simple wrapper.h:

#include <rpm/rpmlib.h>
#include <rpm/rpmts.h>

and then run bindgen on it:

$ CLANG_PATH=/opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/bin/clangLIBCLANG_PATH=/opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/lib/ ~/bin/bindgen wrapper.h --output=rpm.rs

as the result I got these errors:

ERROR:libbindgen::ir::context: Valid declaration with no USR: Cursor(__va_list_tag kind: StructDecl, loc: builtin definitions, usr: None), Some(Cursor(va_list kind: TypedefDecl, loc: /opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/bin/../lib/clang/3.9.1/include/stdarg.h:30:27, usr: Some("c:@T@va_list")))
ERROR:libbindgen::ir::context: Valid declaration with no USR: Cursor(__builtin_va_list kind: TypedefDecl, loc: builtin definitions, usr: None), Some(Cursor(va_list kind: TypedefDecl, loc: /opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/bin/../lib/clang/3.9.1/include/stdarg.h:30:27, usr: Some("c:@T@va_list")))

Do I miss something?

The text was updated successfully, but these errors were encountered:

emilio · 2016-12-27T17:53:52Z

Those errors are non-fatal, does the generated code compile? You can post it here if you want.

If it doesn't, I bet this is #361, and you need to blacklist the functions that use varargs until we do it automatically.

fitzgen · 2016-12-28T17:52:32Z

We might want to make all the non-fatal error!(...)s into warn!(...)s, because I expect most people get confused. I know I did when I first started using bindgen.

Thoughts @emilio ?

emilio · 2016-12-28T18:35:59Z

Sounds good to me.

sa2ajj · 2016-12-28T18:45:25Z

right, since I had no rpm.rs after my previous attempts (for some reason my package installed clang 3.8 was used) I did not check whether the aforementioned file has any content: actually it does.

I tried to compile the produced rpm.rs and it failed: from my inexperienced look, the problems are not related to the issues mentioned in this issue #361. A couple of errors from running rustc on the rpm.rs:

error[E0428]: a type named `poptOption` has already been defined in this module
...
error: unions are unstable and possibly buggy (see issue #32836)
...
error: aborting due to 14 previous errors

and the union related error is the 13 beside the first one.

sa2ajj · 2016-12-28T18:50:15Z

I forgot to mention:

$ rustc --version
rustc 1.14.0 (e8a012324 2016-12-16)

emilio · 2016-12-28T19:24:52Z

Ok, so the unions one is that by default we generate rust unions, which are nightly-only right now. We probably should change that default behavior.

Can you call bindgen with --no-unstable-rust, and see if that makes things better? The poptOption error seems interesting.

sa2ajj · 2016-12-28T19:44:03Z

As you suggested, I run bindgen with --no-unstable-rust option. The compilation of the resulting rpm.rs failed with

error[E0428]: a type named `poptOption` has already been defined in this module
    --> rpm.rs:1246:1
     |   
1176 | pub struct poptOption {
     | - previous definition of `poptOption` here
...
1246 | pub type poptOption = *mut poptOption;
     | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ already defined

error: main function not found

error[E0391]: unsupported cyclic reference between types/traits detected
    --> rpm.rs:1246:28
     |   
1246 | pub type poptOption = *mut poptOption;
     |                            ^^^^^^^^^^ cyclic reference
     |   
     = note: the cycle begins when processing `poptOption`...
     = note: ...which then again requires processing `poptOption`, completing the cycle.

error: aborting due to previous error

main function not found is understandable, the rest is not that much.

poptOption comes from popt library (#include <popt.h>) and it has this:

struct poptOption {
...
};
...
#ifndef __cplusplus
/*@-exporttype -typeuse@*/
typedef struct poptOption * poptOption;
/*@=exporttype =typeuse@*/
#endif
...

Is C++ somehow controlled by an option?

emilio · 2016-12-28T19:54:54Z

Yup! passing -x c++ to bindgen like:

./bindgen wrapper.h --output=rpm.rs --no-unstable-rust -- -x c++

Should make it work.

sa2ajj · 2016-12-28T20:10:01Z

That definitely made a difference. First, the output of bindgen:

ERROR:libbindgen::ir::item: Unhandled cursor kind CXCursorKind(400): Cursor( kind: UnexposedAttr, loc: /usr/lib/gcc/x86_64-linux-gnu/6.2.1/../../../../include/x86_64-linux-gnu/c++/6.2.1/bits/c++config.h:223:43, usr: None)
ERROR:libbindgen::ir::item: Unhandled cursor kind CXCursorKind(400): Cursor( kind: UnexposedAttr, loc: /usr/lib/gcc/x86_64-linux-gnu/6.2.1/../../../../include/x86_64-linux-gnu/c++/6.2.1/bits/c++config.h:227:43, usr: None)
ERROR:libbindgen::ir::item: Unhandled cursor kind CXCursorKind(417): Cursor(default kind: attribute(visibility), loc: /usr/lib/gcc/x86_64-linux-gnu/6.2.1/../../../../include/c++/6.2.1/cstdlib:120:15, usr: None)
ERROR:libbindgen::ir::item: Unhandled cursor kind CXCursorKind(417): Cursor(default kind: attribute(visibility), loc: /usr/lib/gcc/x86_64-linux-gnu/6.2.1/../../../../include/c++/6.2.1/cstdlib:215:21, usr: None)
ERROR:libbindgen::ir::context: Valid declaration with no USR: Cursor(__va_list_tag kind: StructDecl, loc: builtin definitions, usr: None), Some(Cursor(va_list kind: TypedefDecl, loc: /opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/bin/../lib/clang/3.9.1/include/stdarg.h:30:27, usr: Some("c:@T@va_list")))
ERROR:libbindgen::ir::context: Valid declaration with no USR: Cursor(__builtin_va_list kind: TypedefDecl, loc: builtin definitions, usr: None), Some(Cursor(va_list kind: TypedefDecl, loc: /opt/clang+llvm-3.9.1-x86_64-linux-gnu-debian8/bin/../lib/clang/3.9.1/include/stdarg.h:30:27, usr: Some("c:@T@va_list")))

and then the output of rustc:

error: main function not found

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5840:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5840 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5848:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5848 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5854:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5854 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5862:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5862 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5868:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5868 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5876:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5876 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5882:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5882 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5890:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5890 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5896:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5896 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5904:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5904 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5910:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5910 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5918:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5918 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5924:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5924 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5932:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5932 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5938:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5938 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5946:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5946 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5952:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5952 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5960:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5960 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `at_offset_0`:
    --> rpm.rs:5966:5
     |
5826 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     - previous definition of `at_offset_0` here
...
5966 |     pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
     |     ^ duplicate definition

error[E0201]: duplicate definitions with name `set_at_offset_0`:
    --> rpm.rs:5974:5
     |
5834 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     - previous definition of `set_at_offset_0` here
...
5974 |     pub fn set_at_offset_0(&mut self, val: ::std::os::raw::c_int) {
     |     ^ duplicate definition

error: aborting due to 20 previous errors

emilio · 2016-12-28T20:28:06Z

That's... Interesting!

Do you have the definition of the struct that is causing those?

sa2ajj · 2016-12-30T09:36:02Z

this comes from <bits/timex.h>, I believe, which means that bindgen tries to generate for all things system.

what I want is to generate bindings for rpm related headers only, is there an easy way to do that?

the code rustc complains about:

#[repr(C)]
#[derive(Debug, Copy)]
pub struct timex {
    pub modes: ::std::os::raw::c_uint,
    pub offset: __syscall_slong_t,
    pub freq: __syscall_slong_t,
    pub maxerror: __syscall_slong_t,
    pub esterror: __syscall_slong_t,
    pub status: ::std::os::raw::c_int,
    pub constant: __syscall_slong_t,
    pub precision: __syscall_slong_t,
    pub tolerance: __syscall_slong_t,
    pub time: timeval,
    pub tick: __syscall_slong_t,
    pub ppsfreq: __syscall_slong_t,
    pub jitter: __syscall_slong_t,
    pub shift: ::std::os::raw::c_int,
    pub stabil: __syscall_slong_t,
    pub jitcnt: __syscall_slong_t,
    pub calcnt: __syscall_slong_t,
    pub errcnt: __syscall_slong_t,
    pub stbcnt: __syscall_slong_t,
    pub tai: ::std::os::raw::c_int,
    pub _bitfield_1: u32,
    pub _bitfield_2: u32,
    pub _bitfield_3: u32,
    pub _bitfield_4: u32,
    pub _bitfield_5: u32,
    pub _bitfield_6: u32,
    pub _bitfield_7: u32,
    pub _bitfield_8: u32,
    pub _bitfield_9: u32,
    pub _bitfield_10: u32,
    pub _bitfield_11: u32,
}
#[test]
fn bindgen_test_layout_timex() {
    assert_eq!(::std::mem::size_of::<timex>() , 208usize);
    assert_eq!(::std::mem::align_of::<timex>() , 8usize);
}
impl Clone for timex {
    fn clone(&self) -> Self { *self }
}
impl timex {
    #[inline]
    pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
        unsafe {
            ::std::mem::transmute(((self._bitfield_1 &
                                        (4294967295usize as u32)) >> 0u32) as
                                      u32)
        }
    }
    ...
    #[inline]
    pub fn at_offset_0(&self) -> ::std::os::raw::c_int {
        unsafe {
            ::std::mem::transmute(((self._bitfield_2 &
                                        (4294967295usize as u32)) >> 0u32) as
                                      u32)
        }
    }
    ...

emilio · 2016-12-30T11:05:29Z

Yes, you should whitelist the rpm stuff you want, using regexes.

For example: `--whitelist-function "rpm.*".

fitzgen · 2017-07-21T18:15:16Z

@sa2ajj are you still hitting this on master? If so, the best way to contribute to resolving this bug would be to create a reduced test case.

sa2ajj · 2017-10-25T17:15:01Z

sorry for a slow follow-up: i'll check it during the coming weekend :/

tarcieri · 2018-04-09T02:46:04Z

I've managed to use bindgen to produce a Rust binding for librpm.so, librpmio.so, librpmbuild.so, and librpmsign.so (heads up @sa2ajj):

https://github.com/iqlusion-io/crates/tree/master/rpmlib-sys

I should start by noting that all of the issues being discussed in this thread, at least for me, have been fixed, and IMO the issue can probably be closed.

I ran into several other minor (not-really-bindgen-bug) issues, most of which I've managed to address. Many of them were fixed by having bindgen build a C++ binding instead of a C one. If you want the gorey details you can read about them here:

iqlusioninc/crates#11
iqlusioninc/crates#12
iqlusioninc/crates#17

emilio · 2018-04-09T09:14:41Z

Ok, I'll close this based on that, thanks @tarcieri!

rpmlib-sys: rpmts.h and rpmfi.h cause bindgen_test_layout_max_align_t failures iqlusioninc/crates#11 should be fixed with rust_target(1.25), since that supports repr(align).
rpmlib-sys: popt.h causes bindgen to output invalid self-referential poptOption definition iqlusioninc/crates#12 is kind of hard to fix without introducing more complex opt-in name namespacing, but it's tracked on typedef struct X * X generates invalid Rust code #1252.

tony-iqlusion · 2018-04-09T15:09:38Z

@emilio thanks for the pointers... looks like #1252 is about the same silly header file I was having trouble with on iqlusioninc/crates#12 ! Glad I found a workaround.

fitzgen added I-needs-reduced-test-case I-unclear labels Jul 21, 2017

emilio closed this as completed Apr 9, 2018

tarcieri mentioned this issue Apr 9, 2018

rpmlib-sys: rpmts.h and rpmfi.h cause bindgen_test_layout_max_align_t failures iqlusioninc/crates#11

Closed

leo60228 mentioned this issue Jul 21, 2020

Change non-fatal errors to warnings #1833

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

failed to produce bindings for RPM library #363

failed to produce bindings for RPM library #363

sa2ajj commented Dec 27, 2016

emilio commented Dec 27, 2016

fitzgen commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016 •

edited

Loading

sa2ajj commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016 •

edited

Loading

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 30, 2016

emilio commented Dec 30, 2016

fitzgen commented Jul 21, 2017

sa2ajj commented Oct 25, 2017

tarcieri commented Apr 9, 2018 •

edited

Loading

emilio commented Apr 9, 2018

tony-iqlusion commented Apr 9, 2018 •

edited

Loading

failed to produce bindings for RPM library #363

failed to produce bindings for RPM library #363

Comments

sa2ajj commented Dec 27, 2016

emilio commented Dec 27, 2016

fitzgen commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016 • edited Loading

sa2ajj commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016 • edited Loading

emilio commented Dec 28, 2016

sa2ajj commented Dec 28, 2016

emilio commented Dec 28, 2016

sa2ajj commented Dec 30, 2016

emilio commented Dec 30, 2016

fitzgen commented Jul 21, 2017

sa2ajj commented Oct 25, 2017

tarcieri commented Apr 9, 2018 • edited Loading

emilio commented Apr 9, 2018

tony-iqlusion commented Apr 9, 2018 • edited Loading

sa2ajj commented Dec 28, 2016 •

edited

Loading

sa2ajj commented Dec 28, 2016 •

edited

Loading

tarcieri commented Apr 9, 2018 •

edited

Loading

tony-iqlusion commented Apr 9, 2018 •

edited

Loading