cmd/cgo: _Cfunc_CString redeclared in this block

[IoIxD]

Go version

1.22.3

Output of go env in your module/workspace:

GO111MODULE=''
GOARCH='amd64'
GOBIN=''
GOCACHE='/home/gavin/.cache/go-build'
GOENV='/home/gavin/.config/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFLAGS=''
GOHOSTARCH='amd64'
GOHOSTOS='linux'
GOINSECURE=''
GOMODCACHE='/home/gavin/go/pkg/mod'
GONOPROXY=''
GONOSUMDB=''
GOOS='linux'
GOPATH='/home/gavin/go'
GOPRIVATE=''
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/usr/lib/go'
GOSUMDB='sum.golang.org'
GOTMPDIR=''
GOTOOLCHAIN='auto'
GOTOOLDIR='/usr/lib/go/pkg/tool/linux_amd64'
GOVCS=''
GOVERSION='go1.22.3'
GCCGO='gccgo'
GOAMD64='v1'
AR='ar'
CC='gcc'
CXX='g++'
CGO_ENABLED='1'
GOMOD='/home/gavin/Projects/sql-builder-xd/go.mod'
GOWORK=''
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
PKG_CONFIG='pkg-config'
GOGCCFLAGS='-fPIC -m64 -pthread -Wl,--no-gc-sections -fmessage-length=0 -ffile-prefix-map=/tmp/go-build2614403304=/tmp/go-build -gno-record-gcc-switches'

What did you do?

Tried to create a cgo package and use CString.

// #cgo CFLAGS: -g -Wall
// #include <stdlib.h>
// #include "native/sql_builder_rs.h"
import "C"

func FunctionWithString(table string) error {
	st, err := C.CString(table)
	if err != nil {
		return err
	}
	C.external_function(C.new_rust_string(st))
}

What did you see happen?

golang gives an error that CString was defined twice, and then tries to use a CString function that doesn't exist.

# github.com/IoIxD/sql-builder-xd
$WORK/b001/_cgo_gotypes.go:75:6: _Cfunc_CString redeclared in this block
	$WORK/b001/_cgo_gotypes.go:53:6: other declaration of _Cfunc_CString
./bridge.go:9:12: undefined: _C2func_CString
./bridge.go:13:1: missing return

What did you expect to see?

I expected CString to work, and a compiler error to not be given...at least from Golang.
I don't think Rust is relevant here because I actually expected to get a linker error because I haven't linked to the Rust library yet, I'm just including the header file.

It might also be that I'm on Arch Linux, a decision that I regret. If Arch is at fault here, I am willing to switch to Fedora.

[ianlancetaylor]

Can you show us a copy of native/sql_builder_rs.h and any files that it might include? Thanks.

[IoIxD]

Can you show us a copy of native/sql_builder_rs.h and any files that it might include? Thanks.

https://hastebin.com/share/unomecitik.scss

It's a bit weird due to being generated by Rust's cbindgen.

[ianlancetaylor]

Thanks. Unfortunately that didn't really help. Is there a way that we can recreate the problem ourselves?

[IoIxD]

I have the repo right here: https://github.com/IoIxD/sql-builder-xd/
The repo includes the .a file to link to, just do go run .

[ianlancetaylor]

Thanks. When I clone and build that repo on linux-amd64, I get

 github.com/IoIxD/sql-builder-xd
./bridge.go:16:25: cannot use st (variable of type *_Ctype_char) as *_Ctype_uchar value in argument to (_Cfunc_new_rust_string)

When I change the call in bridge.go to

	h := C.new_rust_string((*C.uint8_t)(unsafe.Pointer(st)))

the code builds without error.

So I am unable to reproduce the problem.

[IoIxD]

Hi, do you know how I can remove all the golang files to do a clean reinstall? I don't think my package manager does this correctly.

[ianlancetaylor]

Sorry, I don't know. It really depends on how you installed it in the first place.

[IoIxD]

I opened up my laptop and applied your fix and the bug fixed itself.

[IoIxD]

Further work leads me to realize that this is actually an error message that is caused by other error messages. The Go compiler is misreporting the real error message, which in the original case was the *C.uint8_t problem.

Before, the IDE WAS telling me that error, and it didn't show up when compiling so I thought it was a fluke. This time, I don't even get that.

func (*SqlBuilderStart) SelectValues(values []string) *SqlBuilder {
	vals := make([]*C.int8_t, len(values))
	for i, v := range values {
		st, _ := C.CString(v)
		vals[i] = (*C.int8_t)(st)
	}
	s := C.new_sql_builder_select_values((**C.int8_t)(&vals[0]), (C.ulong)(len(values)))
	for _, v := range vals {
		C.free(unsafe.Pointer(v))

	}
	return &SqlBuilder{s}
}

Funnily enough, the error also only appears on the new usage of C.CString, on line 32.

[IoIxD]

Oh my god I got the actual error.

A previous error had told me that C.CString returns a message and an error. This is false, but when I assume it returns an error, the error appears.