link modifier export-symbols: export all global symbols from selected uptream c static libraries

[cezarbbb]

In order to be able to export symbols from a specified upstream C static library, I redesigned a solution that, compared to a previous PR #150335 I submitted, will not have any extra symbols leaking out.

The following points should be noted:

• This attribute will select and import the Global symbols of the first matching library it finds.
• Developers should ensure that there are no libraries with the same name.
• This modifier is only compatible with static linking kind
• By default, upstream C static libraries will not export their Global symbols regardless of whether LTO optimization is enabled. However, after enabling this attribute, if the upstream C static library has LTO optimization enabled, the compiler will issue an error to inform the developer that the linked C library is invalid.

The test code is the same as the PR #150335.
Here are the results:

1. cargo +include-libs rustc --release -- -L. -lstatic:+export-symbols=c_add
   (or you can use #[link(name = "c_add", kind= "static", modifier = "+export-symbols")] in the file)

                 U abort@GLIBC_2.2.5
                 U bcmp@GLIBC_2.2.5
0000000000014f60 T c_add
                 U calloc@GLIBC_2.2.5
                 U close@GLIBC_2.2.5
0000000000014f70 T c_sub
                 w __cxa_finalize@GLIBC_2.2.5
                 w __cxa_thread_atexit_impl@GLIBC_2.18
                 U dl_iterate_phdr@GLIBC_2.2.5
0000000000014ee0 T downstream_add
                 U __errno_location@GLIBC_2.2.5
                 U free@GLIBC_2.2.5
                 U fstat64@GLIBC_2.33
                 U getcwd@GLIBC_2.2.5
                 U getenv@GLIBC_2.2.5
                 w __gmon_start__
                 w _ITM_deregisterTMCloneTable
                 w _ITM_registerTMCloneTable
                 U lseek64@GLIBC_2.2.5
                 U malloc@GLIBC_2.2.5
                 U memcpy@GLIBC_2.14
                 U memmove@GLIBC_2.2.5
                 U memset@GLIBC_2.2.5
                 U mmap64@GLIBC_2.2.5
                 U munmap@GLIBC_2.2.5
                 U open64@GLIBC_2.2.5
                 U posix_memalign@GLIBC_2.2.5
                 U pthread_key_create@GLIBC_2.34
                 U pthread_key_delete@GLIBC_2.34
                 U pthread_setspecific@GLIBC_2.34
                 U read@GLIBC_2.2.5
                 U readlink@GLIBC_2.2.5
                 U realloc@GLIBC_2.2.5
                 U realpath@GLIBC_2.3
                 U stat64@GLIBC_2.33
                 w statx@GLIBC_2.28
                 U strlen@GLIBC_2.2.5
                 U syscall@GLIBC_2.2.5
                 U __tls_get_addr@GLIBC_2.3
                 U _Unwind_Backtrace@GCC_3.3
                 U _Unwind_DeleteException@GCC_3.0
                 U _Unwind_GetDataRelBase@GCC_3.0
                 U _Unwind_GetIP@GCC_3.0
                 U _Unwind_GetIPInfo@GCC_4.2.0
                 U _Unwind_GetLanguageSpecificData@GCC_3.0
                 U _Unwind_GetRegionStart@GCC_3.0
                 U _Unwind_GetTextRelBase@GCC_3.0
                 U _Unwind_RaiseException@GCC_3.0
                 U _Unwind_Resume@GCC_3.0
                 U _Unwind_SetGR@GCC_3.0
                 U _Unwind_SetIP@GCC_3.0
                 U write@GLIBC_2.2.5
                 U writev@GLIBC_2.2.5

3. cargo +nightly rustc --release -- -L ./

                 U abort@GLIBC_2.2.5
                 U bcmp@GLIBC_2.2.5
                 U calloc@GLIBC_2.2.5
                 U close@GLIBC_2.2.5
                 w __cxa_finalize@GLIBC_2.2.5
                 w __cxa_thread_atexit_impl@GLIBC_2.18
                 U dl_iterate_phdr@GLIBC_2.2.5
0000000000011e10 T downstream_add
                 U __errno_location@GLIBC_2.2.5
                 U free@GLIBC_2.2.5
                 U fstat64@GLIBC_2.33
                 U getcwd@GLIBC_2.2.5
                 U getenv@GLIBC_2.2.5
                 w gettid@GLIBC_2.30
                 w __gmon_start__
                 w _ITM_deregisterTMCloneTable
                 w _ITM_registerTMCloneTable
                 U lseek64@GLIBC_2.2.5
                 U malloc@GLIBC_2.2.5
                 U memcpy@GLIBC_2.14
                 U memmove@GLIBC_2.2.5
                 U memset@GLIBC_2.2.5
                 U mmap64@GLIBC_2.2.5
                 U munmap@GLIBC_2.2.5
                 U open64@GLIBC_2.2.5
                 U posix_memalign@GLIBC_2.2.5
                 U pthread_key_create@GLIBC_2.34
                 U pthread_key_delete@GLIBC_2.34
                 U pthread_setspecific@GLIBC_2.34
                 U read@GLIBC_2.2.5
                 U readlink@GLIBC_2.2.5
                 U realloc@GLIBC_2.2.5
                 U realpath@GLIBC_2.3
                 U stat64@GLIBC_2.33
                 w statx@GLIBC_2.28
                 U strlen@GLIBC_2.2.5
                 U syscall@GLIBC_2.2.5
                 U __tls_get_addr@GLIBC_2.3
                 U _Unwind_Backtrace@GCC_3.3
                 U _Unwind_GetDataRelBase@GCC_3.0
                 U _Unwind_GetIP@GCC_3.0
                 U _Unwind_GetIPInfo@GCC_4.2.0
                 U _Unwind_GetLanguageSpecificData@GCC_3.0
                 U _Unwind_GetRegionStart@GCC_3.0
                 U _Unwind_GetTextRelBase@GCC_3.0
                 U _Unwind_RaiseException@GCC_3.0
                 U _Unwind_Resume@GCC_3.0
                 U _Unwind_SetGR@GCC_3.0
                 U _Unwind_SetIP@GCC_3.0
                 U write@GLIBC_2.2.5
                 U writev@GLIBC_2.2.5

r? @bjorn3

[bjorn3]

We do need to distinguish between text and data symbols. Also on Windows I believe and on macOS I know we currently use the pre-C mangling names in exported_symbols. I think the post mangling name is fine here too for Windows, but you did need to check. On macOS however it is definitively not ok. The symbol names in the object file are prefixed with _, but the linker doesn't accept this.

[cezarbbb]

After testing, the symbol does have a prefix _ on macos, but not on windows. I have removed the prefix symbol (I don’t know why it was added back in the export_symbols function?)

[mati865]

64-bit Windows symbols do not use special mangling, but 32-bit symbols are prefixed with an additional leading underscore.

So:

int foo() { return 42; }

Would be foo symbol on 64-bit Windows, and _foo on 32-bit Windows.

[bjorn3]

And extern "stdcall" has extra mangling, right?

[mati865]

Not only that, all calling conventions have their own mangling: https://learn.microsoft.com/en-us/cpp/error-messages/tool-errors/name-decoration?view=msvc-170

I don't remember whether linkers can export symbols using .DEF files if the symbol doesn't exactly match calling convention. IIRC, GNU ld and LLD (not so sure about this one) can handle it, at least in some cases.

[cezarbbb]

OMG. This is an area I'm unfamiliar with, and I don't have a 32-bit Windows device to test on. I'm unsure how to handle these symbols in 32-bit Windows; perhaps the linker will handle them correctly? Because I can't seem to find any special code in rustc that handles these symbols.

[mati865]

You can test it on 64-bit Windows or any other system by cross-compiling to a i686 Windows target.

Anyway, once you add the test to this repo CI will be able to test it on all systems.

[cezarbbb]

I left a fixme, and I would like to provide the handling of 32-bit Windows as a future PR, is that ok?

[bjorn3]

I think this is a reasonable approach. This will need a test though.

[cezarbbb]

I think this is a reasonable approach. This will need a test though.

I've add a simple test.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[cezarbbb]

@rustbot ready

[petrochenkov]

CrateInfo already has used_libraries (or native_libraries) which can be filtered by export_symbols: Some(true) in link.rs, this field shouldn't be necessary.

[petrochenkov]

The search should be done using find_native_static_library.

Ideally NativeLib::filename would be used, but it's currently set only for bundled libraries, and changing that may mess up other things.