CGO segmentation violation

[Luke Mauldin]

All,

I have a Go program using CGO to access the libvnc library and I am getting this signal and stack trace below.  Even though the stack trace shows it is the C function rfbProcessEvents where the crash is occurring, how can I get more information about the C part of the stack trace?  The libvnc libraries are compiled with debug information and if I call them from a C program and the C program crashes then the crash output will tell me exactly what line is causing the crash but Go doesn't give me that information.  Can I configure the Go program to generate a core dump that would be examinable by GDB that would contain additional information about the crash?  Is there another method that I can use to obtain more information?

SIGSEGV: segmentation violation

PC=0x7f70308821e4

signal arrived during cgo execution

runtime.cgocall(0x402040, 0x7f702c064f68)

/opt/go/src/pkg/runtime/cgocall.c:149 +0x11b fp=0x7f702c064f50

tf/vncmulticastserver/vncmulticast._Cfunc_rfbProcessEvents(0x7f6ffc000920, 0x1f4, 0x134397756a5b2c00)

tf/vncmulticastserver/vncmulticast/_obj/_cgo_defun.c:172 +0x31 fp=0x7f702c064f68

tf/vncmulticastserver/vncmulticast.(*MulticastSession).runServer(0xc2100634e0)

/opt/gopath/src/tf/vncmulticastserver/vncmulticast/multicastsession.go:316 +0xeb fp=0x7f702c064f98

runtime.goexit()

/opt/go/src/pkg/runtime/proc.c:1394 fp=0x7f702c064fa0

created by tf/vncmulticastserver/vncmulticast.(*MulticastSession).initializeServer

/opt/gopath/src/tf/vncmulticastserver/vncmulticast/multicastsession.go:295 +0x29c

[Ian Lance Taylor]

On Fri, Dec 27, 2013 at 5:32 AM, Luke Mauldin <lukem...@gmail.com> wrote:
>
> I have a Go program using CGO to access the libvnc library and I am getting
> this signal and stack trace below. Even though the stack trace shows it is
> the C function rfbProcessEvents where the crash is occurring, how can I get
> more information about the C part of the stack trace? The libvnc libraries
> are compiled with debug information and if I call them from a C program and
> the C program crashes then the crash output will tell me exactly what line
> is causing the crash but Go doesn't give me that information. Can I
> configure the Go program to generate a core dump that would be examinable by
> GDB that would contain additional information about the crash?

You should be able to get a core dump if you set the environment
variable GOTRACEBACK=crash. Unfortunately you may run into
http://golang.org/issue/6988 , which is fixed on tip but not in the
1.2 release.

> Is there
> another method that I can use to obtain more information?

Can you run the program under gdb?

Ian

[Luke Mauldin]

Thank you for the helpful information.  If I apply the patch to fix issue 6988, and set the environment variables, then where will the Go program put the coredump files on Linux after the crash?

[Ian Lance Taylor]

On Fri, Dec 27, 2013 at 6:38 AM, Luke Mauldin <lukem...@gmail.com> wrote:
>
> Thank you for the helpful information. If I apply the patch to fix issue
> 6988, and set the environment variables, then where will the Go program put
> the coredump files on Linux after the crash?

The core dump is really generated by the kernel. I think it will go
into the program's current directory. So don't use "go run" to run
the program--unless you take extra steps, that will run it in a
temporary directory that will then be deleted. Instead, use "go
build" and then run the program.

Ian

[Ken Allen]

You can change kernel.core_pattern with sysctl to change where the core file goes if you need to.

[Luke Mauldin]

I applied the patch for issue 6988 and Go did generate a core dump.  However, when I am using GDB to inspect it I am not getting the expected results.  When I load the core dump file in GDB I get the stack trace below but as you can see from the crash output the program is crashing in the C function "rfbScreenCleanup".  That C function comes from the libvnc library and the library is compiled with debugging information and I would like to use GDB to see exactly what line in the rfbScreenCleanup function caused the crash.  Is that possible?

Luke

GDB output:

(gdb) bt

#0  runtime.raise (sig=void) at /opt/go/src/pkg/runtime/sys_linux_amd64.s:86

#1  0x000000000041e526 in runtime.crash ()

#2  0x000000000041e15f in runtime.sighandler (sig=void, info=void, ctxt=void, gp=void) at /opt/go/src/pkg/runtime/signal_amd64.c:132

#3  0x0000000000428956 in runtime.sigtramp () at /opt/go/src/pkg/runtime/sys_linux_amd64.s:209

#4  0x0000000000428970 in ?? () at /opt/go/src/pkg/runtime/sys_linux_amd64.s:215

#5  0x0000000000000001 in ?? ()

#6  0x0000000000000000 in ?? ()

Crash output:

SIGSEGV: segmentation violation

PC=0x7f4644056460

signal arrived during cgo execution

runtime.cgocall(0x4020a0, 0x7f46645a4ec8)

/opt/go/src/pkg/runtime/cgocall.c:149 +0x11b fp=0x7f46645a4eb0

tf/vncmulticastserver/vncmulticast._Cfunc_rfbScreenCleanup(0x7f46500008c0)

tf/vncmulticastserver/vncmulticast/_obj/_cgo_defun.c:181 +0x31 fp=0x7f46645a4ec8

tf/vncmulticastserver/vncmulticast.(*MulticastSession).Close(0xc2100d79c0, 0x1)

/opt/gopath/src/tf/vncmulticastserver/vncmulticast/multicastsession.go:86 +0x103 fp=0x7f46645a4f28

tf/vncmulticastserver/vncmulticast.(*MulticastSession).listenShutdown(0xc2100d79c0)

/opt/gopath/src/tf/vncmulticastserver/vncmulticast/multicastsession.go:334 +0x14e fp=0x7f46645a4f98

runtime.goexit()

/opt/go/src/pkg/runtime/proc.c:1394 fp=0x7f46645a4fa0

created by tf/vncmulticastserver/vncmulticast.NewMulticastSession

/opt/gopath/src/tf/vncmulticastserver/vncmulticast/multicastsession.go:349 +0xe8

[andrey mirtchovski]

try "thread apply all bt"

> --
> You received this message because you are subscribed to the Google Groups
> "golang-nuts" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to golang-nuts...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.