Re: gdb 5.3 bug

[Daniel Jacobowitz <drow at mvista dot com>]

On Sat, Mar 01, 2003 at 06:59:45PM +0300, Lev Assinovsky wrote:
> Dear Daniel!
> Thank you very match for your advise!
> Snapshot gdb+dejagnu-20030228 works just fine!
> Thank a lot all guys who is involved in gdb development. 
> I am sorry but I have to bother you with another gdb problem.
> The problem is I can't step into the function through the pointer
> if the function resides in dynamically loaded shared object (library).
> Here is the simple test case:

Hmm.  Your testcase works on GNU/Linux, so it must be related to
i386-solaris shared library support.  Maybe someone else on the list
can help.

> 
> 1. Shared object:
> ---------- cut here --------
> #include <iostream>
> using namespace std;
> 
> extern "C" {
>     void my_func()
>     {
>         cout << "This is my_func" << endl;
>     }
> 
> }
> ---------- cut here --------
> 2. Main:
> #include <iostream>
> #include <memory>
> #include <dlfcn.h>
> #include <link.h>
> 
> using namespace std;
> 
> void * handle;
> 
> typedef void  ( *Func_t)();
> 
> Func_t getFunc() 
> {
>     Func_t p = (Func_t)dlsym ( handle, "my_func" );
>     if ( NULL == p )
>     {
>         cout << dlerror() << endl;
>         exit(2);
>     }
>     return p;
> }
> 
> int main()
> {
>     handle = (void*) ::dlopen("./func_so/bin/i386-sun-solaris/func_so.so", RTLD_NOW | RTLD_LOCAL | RTLD_GROUP);
>     if (handle == NULL)
>     {
>         cout << dlerror() << endl;
>         exit(2);
>     }
> 
>     Func_t f = getFunc();
>     f();
> }
> ---------- cut here --------
> 
> This testcase works, but I can't step into f() (last line) in gdb.
> Is it possible to work around or fix this problem?
> Thanks in advance!
> 
> ----
> Lev Assinovsky
> Aelita Software Corporation
> O&S Core Division, Programmer
> ICQ# 165072909
> 
> 
> > -----Original Message-----
> > From: Daniel Jacobowitz [mailto:drow at mvista dot com]
> > Sent: Friday, February 28, 2003 8:19 PM
> > To: Lev Assinovsky; gdb at sources dot redhat dot com
> > Subject: Re: gdb 5.3 bug
> > 
> > 
> > On Thu, Feb 27, 2003 at 02:09:30PM -0500, Daniel Jacobowitz wrote:
> > > On Thu, Feb 27, 2003 at 09:11:14PM +0300, Lev Assinovsky wrote:
> > > > GNU gdb 5.3
> > > > Copyright 2002 Free Software Foundation, Inc.
> > > > GDB is free software, covered by the GNU General Public 
> > License, and you are
> > > > welcome to change it and/or distribute copies of it under 
> > certain conditions.
> > > > Type "show copying" to see the conditions.
> > > > There is absolutely no warranty for GDB.  Type "show 
> > warranty" for details.
> > > > This GDB was configured as "i386-pc-solaris2.8"...
> > > > (gdb) l
> > > > 
> > > > Program received signal SIGSEGV, Segmentation fault.
> > > > 0x08136671 in finish_block (symbol=0x6e5f5f3a, 
> > listhead=0x82241f4, old_blocks=0x3a787863, start=1634562671, 
> > > >     end=1634562720, objfile=0x8263fe0) at buildsym.c:304
> > > > 304           struct type *ftype = SYMBOL_TYPE (symbol);
> > > > (gdb) 
> > > > 
> > > > 
> > > > Any clue what might be the reason of gdb crash?
> > > 
> > > OK, the value of "symbol" is obviously wrong.  Could you 
> > privately send
> > > me the binary?  If that's not possible, at least a 
> > backtrace would be
> > > useful...
> > 
> > 
> > Lev,
> > 
> > You're in luck, this has already been fixed.  It appears to 
> > be a bug in
> > the Sun compilers - it outputs end-of-function markers without a
> > corresponding beginning-of-function marker.  If you get a GDB snapshot
> > from CVS (http://sources.redhat.com/gdb/) it should handle 
> > your program
> > gracefully.
> > 
> > -- 
> > Daniel Jacobowitz
> > MontaVista Software                         Debian GNU/Linux Developer
> > 
> 

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer