如何释放Rust中通过FFI分配的*字符?
我通过Rust的FFI调用LLVM API<代码>LLVMPrintModuleToString。但是,当我将指针包装在如何释放Rust中通过FFI分配的*字符?,rust,ffi,Rust,Ffi,我通过Rust的FFI调用LLVM APILLVMPrintModuleToString。但是,当我将指针包装在CString中时,如果锈掉了指针,就会出现错误 #![feature(cstr_memory)] use std::ffi::CString; extern crate llvm_sys as llvm; fn main() { let llvm_ir_cstring; unsafe { let module = llvm::core::LLVMM
CString
中时,如果锈掉了指针,就会出现错误
#![feature(cstr_memory)]
use std::ffi::CString;
extern crate llvm_sys as llvm;
fn main() {
let llvm_ir_cstring;
unsafe {
let module = llvm::core::LLVMModuleCreateWithName(b"nop\0".as_ptr() as *const _);
let llvm_ir_char_ptr = llvm::core::LLVMPrintModuleToString(module);
llvm::core::LLVMDisposeModule(module);
llvm_ir_cstring = CString::from_ptr(llvm_ir_char_ptr);
}
println!("llvm_ir_cstring: {:?}", llvm_ir_cstring);
}
Valgrind错误:
==10413== Invalid read of size 4
==10413== at 0x4870586: pthread_mutex_lock (in /usr/lib/libpthread-2.21.so)
==10413== by 0x17F89B: je_arena_dalloc_small (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x178C30: je_sdallocx (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x10FA57: heap::imp::deallocate::h1fb92d59333c497bkja (heap.rs:268)
==10413== by 0x10F999: heap::deallocate::h5680e3fedc9e96320da (heap.rs:89)
==10413== by 0x10F929: heap::exchange_free::h452463f962f7ec23kfa (heap.rs:131)
==10413== by 0x10F8C5: Box$LT$$u5b$u8$u5d$$GT$::drop.1769::haf7017472635c7cf (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x10F836: std..ffi..c_str..CString::drop.1766::h04d2b3db8d468f0c (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x10F5FF: main::h04b7feb343e229ccgaa (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x16DBCA: rt::unwind::try::try_fn::h2403085009213705177 (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x16FF5A: rust_try_inner (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== by 0x16FF33: rust_try (in /home/wilfred/projects/bfc/target/debug/bfc)
==10413== Address 0x1d684 is not stack'd, malloc'd or (recently) free'd
为什么会这样?处理C API中的*char
的正确方法是什么?根据:
使用LLVMDisposeMessage释放字符串
在一般情况下,如果调用库中分配内存的函数,则应调用该库中释放内存的另一个函数;这通常应记录为职能部门合同的一部分
如果函数的文档告诉您使用free
,那么如果您的应用程序没有链接到与库的malloc
对应的free
(例如,您的应用程序链接到msvcr120,但库链接到msvcr100),则会出现问题。这就是为什么好的库提供了一种方法来释放它为您分配的资源
Rust中的默认内存分配器不是C的
malloc
,而是另一个调用的分配器CString
假设字符串是用Rust的内存分配器分配的,因此当CString
的析构函数运行时,它会运行jemalloc的free
(您可以从调用堆栈中的je_
前缀函数看出),但是它失败了,因为字符串没有分配给jemalloc的malloc
让一个C函数释放它,并调用这个C函数?谢谢,这真的很有启发性。