cgo panic stack trace with no proper line number for c code
569 views
Skip to first unread message
andy....@gmail.com
Jan 19, 2016, 8:34:30 AM1/19/16
to golang-nuts
I have recently come across a situation that when processing cgo. When i make a SIGBUS signal to the cgo program, the stack trace come to an error line number pointing to the c code. How does the line number with cgo to be interpreted correctly?
i.e. what is the meaning of "54" after " /home/demo/code/go/src/github.com/demo/a/".
i.e. what is the meaning of "54" after " /home/demo/code/go/src/github.com/demo/a/".
[signal 0x7 code=0x2 addr=0x7f45fe055000 pc=0x36f5889131]runtime stack:runtime.gothrow(0x95b570, 0x2a)/usr/local/go/src/runtime/panic.go:503 +0x8eruntime.sigpanic()/usr/local/go/src/runtime/sigpanic_unix.go:14 +0x5egoroutine 13 [syscall, locked to thread]:runtime.cgocall_errno(0x403aa0, 0xc65d0c1a60, 0x0)/usr/local/go/src/runtime/cgocall.go:130 +0xf5 fp=0xc65d0c1a40 sp=0xc65d0c1a18github.com/demo/a._Cfunc_create(0x7f4614000ac0, 0x12c, 0x568c8d61, 0xc20000000e, 0xc2db7b82a0, 0x0)/home/demo/code/go/src/github.com/demo/a/:54 +0x44 fp=0xc65d0c1a60 sp=0xc65d0c1a40
Ian Lance Taylor
Jan 19, 2016, 1:18:42 PM1/19/16
to andy....@gmail.com, golang-nuts
On Tue, Jan 19, 2016 at 4:23 AM, <andy....@gmail.com> wrote:
> I have recently come across a situation that when processing cgo. When i
> make a SIGBUS signal to the cgo program, the stack trace come to an error
> line number pointing to the c code. How does the line number with cgo to be
> interpreted correctly?
>
> i.e. what is the meaning of "54" after "
> /home/demo/code/go/src/github.com/demo/a/".
That looks strange. I don't know why there is no file name there.
> I have recently come across a situation that when processing cgo. When i
> make a SIGBUS signal to the cgo program, the stack trace come to an error
> line number pointing to the c code. How does the line number with cgo to be
> interpreted correctly?
>
> i.e. what is the meaning of "54" after "
> /home/demo/code/go/src/github.com/demo/a/".
What I do know is that it doesn't matter, in that knowing the answer
won't help. That file is a file generated by cgo to implement the
call to C.create. The actual SIGBUS signal is coming from C code, and
the Go traceback is not showing that code. With Go 1.6 you will need
to run the program under a debugger, or use nm/objdump/readelf to
translate the PC address (0x36f5889131) back into a symbol.
If you feel like experimenting, and you are using amd64 GNU/Linux, you
could try using https://golang.org/cl/17761 in conjunction with the
github.com/ianlancetaylor/cgosymbolize package, to get a symbolic
backtrace of your cgo crash. That is strictly experimental, though,
and not for the unwary. I hope to include it in the 1.7 release.
Ian
andy.xning
Jan 19, 2016, 9:05:32 PM1/19/16
to Ian Lance Taylor, golang-nuts
> What I do know is that it doesn't matter, in that knowing the answer
> won't help. That file is a file generated by cgo to implement the
> call to C.create. The actual SIGBUS signal is coming from C code, and
> the Go traceback is not showing that code.
Does it mean that by default cgo stack trace will not show the exact line number and c file where the error occures? Or, it is only a special case in mine that the exact line number and c file name is losing.
Currently the debug is difficult for a cgo stack trace. :(
原始邮件
发件人: Ian Lance Taylor<ia...@golang.org>
收件人: andy.xning<andy....@gmail.com>
抄送: golang-nuts<golan...@googlegroups.com>
发送时间: 2016年1月20日(周三) 02:18
主题: Re: [go-nuts] cgo panic stack trace with no proper line number for c code
Ian Lance Taylor
Jan 19, 2016, 9:08:17 PM1/19/16
to andy.xning, golang-nuts
On Tue, Jan 19, 2016 at 6:05 PM, andy.xning <andy....@gmail.com> wrote:
>> What I do know is that it doesn't matter, in that knowing the answer
>> won't help. That file is a file generated by cgo to implement the
>> call to C.create. The actual SIGBUS signal is coming from C code, and
>> the Go traceback is not showing that code.
>
>
> Does it mean that by default cgo stack trace will not show the exact line
> number and c file where the error occures? Or, it is only a special case in
> mine that the exact line number and c file name is losing.
The cgo stack trace is not being displayed at all. What is being
>> What I do know is that it doesn't matter, in that knowing the answer
>> won't help. That file is a file generated by cgo to implement the
>> call to C.create. The actual SIGBUS signal is coming from C code, and
>> the Go traceback is not showing that code.
>
>
> Does it mean that by default cgo stack trace will not show the exact line
> number and c file where the error occures? Or, it is only a special case in
> mine that the exact line number and c file name is losing.
displayed is the Go stack trace up to the point of the call into C.
The stack trace of the code in C does not appear.
> Currently the debug is difficult for a cgo stack trace. :(
won't help you today.
Ian
andy.xning
Jan 19, 2016, 9:09:30 PM1/19/16
to Ian Lance Taylor, golang-nuts
Thanks for the info about changing it in 1.7. :)
发件人: Ian Lance Taylor<ia...@golang.org>
收件人: andy.xning<andy....@gmail.com>
抄送: golang-nuts<golan...@googlegroups.com>
发送时间: 2016年1月20日(周三) 10:08
主题: Re: [go-nuts] cgo panic stack trace with no proper line number for c code
Reply all
Reply to author
Forward
0 new messages