RFC: DW_TAG_try_block

RFC: DW_TAG_try_block

[H. J. Lu]

Intel C++ compiler generates DW_TAG_try_block for

	try
	  {
	  ...
	  }

Gdb can't print any local variables declared inside the try block. This
patch seems to work for us. Is that correct?


H.J.
---
2003-05-26  H.J. Lu <hongjiu.lu@intel.com>

	* dwarf2read.c (process_die): Handle DW_TAG_try_block.

--- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
+++ gdb/dwarf2read.c	2003-05-26 12:37:08.000000000 -0700
@@ -1768,6 +1768,7 @@ process_die (struct die_info *die, struc
          of a function and make GDB `next' properly over inlined functions.  */
       break;
     case DW_TAG_lexical_block:
+    case DW_TAG_try_block:
       read_lexical_block_scope (die, objfile, cu_header);
       break;
     case DW_TAG_class_type:

Re: RFC: DW_TAG_try_block

[Daniel Jacobowitz]

On Mon, May 26, 2003 at 12:44:43PM -0700, H. J. Lu wrote:
> Intel C++ compiler generates DW_TAG_try_block for
> 
> 	try
> 	  {
> 	  ...
> 	  }
> 
> Gdb can't print any local variables declared inside the try block. This
> patch seems to work for us. Is that correct?

I believe that this is correct, and that (at least for now) the same
thing should be done for DW_TAG_catch_block.  There's a PR for this
issue.

[But wait for a DWARF maintainer to comment, please.]

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: RFC: DW_TAG_try_block

[H. J. Lu]

On Mon, May 26, 2003 at 03:56:46PM -0400, Daniel Jacobowitz wrote:
> On Mon, May 26, 2003 at 12:44:43PM -0700, H. J. Lu wrote:
> > Intel C++ compiler generates DW_TAG_try_block for
> > 
> > 	try
> > 	  {
> > 	  ...
> > 	  }
> > 
> > Gdb can't print any local variables declared inside the try block. This
> > patch seems to work for us. Is that correct?
> 
> I believe that this is correct, and that (at least for now) the same
> thing should be done for DW_TAG_catch_block.  There's a PR for this
> issue.
> 

You are right. We do use DW_TAG_catch_block. This also works for the
catch block.


H.J.
---
2003-05-26  H.J. Lu <hongjiu.lu@intel.com>

	* dwarf2read.c (process_die): Handle DW_TAG_try_block and
	DW_TAG_catch_block.

--- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
+++ gdb/dwarf2read.c	2003-05-26 13:07:47.000000000 -0700
@@ -1768,6 +1768,8 @@ process_die (struct die_info *die, struc
          of a function and make GDB `next' properly over inlined functions.  */
       break;
     case DW_TAG_lexical_block:
+    case DW_TAG_try_block:
+    case DW_TAG_catch_block:
       read_lexical_block_scope (die, objfile, cu_header);
       break;
     case DW_TAG_class_type:

Re: RFC: DW_TAG_try_block

[Elena Zannoni]

H. J. Lu writes:
 > Intel C++ compiler generates DW_TAG_try_block for
 > 
 > 	try
 > 	  {
 > 	  ...
 > 	  }
 > 
 > Gdb can't print any local variables declared inside the try block. This
 > patch seems to work for us. Is that correct?
 > 
 > 
 > H.J.
 > ---
 > 2003-05-26  H.J. Lu <hongjiu.lu@intel.com>
 > 
 > 	* dwarf2read.c (process_die): Handle DW_TAG_try_block.
 > 
 > --- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
 > +++ gdb/dwarf2read.c	2003-05-26 12:37:08.000000000 -0700
 > @@ -1768,6 +1768,7 @@ process_die (struct die_info *die, struc
 >           of a function and make GDB `next' properly over inlined functions.  */
 >        break;
 >      case DW_TAG_lexical_block:
 > +    case DW_TAG_try_block:
 >        read_lexical_block_scope (die, objfile, cu_header);
 >        break;
 >      case DW_TAG_class_type:


Seems ok, can you address Daniel's concern about the catch block?
I.e. add that as well? 
How about adding a little test case to the gdb.c++ directory?

elena

Re: RFC: DW_TAG_try_block

[H. J. Lu]

On Tue, May 27, 2003 at 10:47:03AM -0400, Elena Zannoni wrote:
> H. J. Lu writes:
>  > Intel C++ compiler generates DW_TAG_try_block for
>  > 
>  > 	try
>  > 	  {
>  > 	  ...
>  > 	  }
>  > 
>  > Gdb can't print any local variables declared inside the try block. This
>  > patch seems to work for us. Is that correct?
>  > 
>  > 
>  > H.J.
>  > ---
>  > 2003-05-26  H.J. Lu <hongjiu.lu@intel.com>
>  > 
>  > 	* dwarf2read.c (process_die): Handle DW_TAG_try_block.
>  > 
>  > --- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
>  > +++ gdb/dwarf2read.c	2003-05-26 12:37:08.000000000 -0700
>  > @@ -1768,6 +1768,7 @@ process_die (struct die_info *die, struc
>  >           of a function and make GDB `next' properly over inlined functions.  */
>  >        break;
>  >      case DW_TAG_lexical_block:
>  > +    case DW_TAG_try_block:
>  >        read_lexical_block_scope (die, objfile, cu_header);
>  >        break;
>  >      case DW_TAG_class_type:
> 
> 
> Seems ok, can you address Daniel's concern about the catch block?
> I.e. add that as well? 

http://sources.redhat.com/ml/gdb/2003-05/msg00348.html

> How about adding a little test case to the gdb.c++ directory?
> 

I can provide a simple C++ code. But you will need Intel C++ compiler to
see the problem. Also the current Intel compiler doesn't support RedHat
9.


H.J.

Re: RFC: DW_TAG_try_block

[Daniel Jacobowitz]

On Tue, May 27, 2003 at 07:57:36AM -0700, H. J. Lu wrote:
> On Tue, May 27, 2003 at 10:47:03AM -0400, Elena Zannoni wrote:
> > H. J. Lu writes:
> >  > Intel C++ compiler generates DW_TAG_try_block for
> >  > 
> >  > 	try
> >  > 	  {
> >  > 	  ...
> >  > 	  }
> >  > 
> >  > Gdb can't print any local variables declared inside the try block. This
> >  > patch seems to work for us. Is that correct?
> >  > 
> >  > 
> >  > H.J.
> >  > ---
> >  > 2003-05-26  H.J. Lu <hongjiu.lu@intel.com>
> >  > 
> >  > 	* dwarf2read.c (process_die): Handle DW_TAG_try_block.
> >  > 
> >  > --- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
> >  > +++ gdb/dwarf2read.c	2003-05-26 12:37:08.000000000 -0700
> >  > @@ -1768,6 +1768,7 @@ process_die (struct die_info *die, struc
> >  >           of a function and make GDB `next' properly over inlined functions.  */
> >  >        break;
> >  >      case DW_TAG_lexical_block:
> >  > +    case DW_TAG_try_block:
> >  >        read_lexical_block_scope (die, objfile, cu_header);
> >  >        break;
> >  >      case DW_TAG_class_type:
> > 
> > 
> > Seems ok, can you address Daniel's concern about the catch block?
> > I.e. add that as well? 
> 
> http://sources.redhat.com/ml/gdb/2003-05/msg00348.html
> 
> > How about adding a little test case to the gdb.c++ directory?
> > 
> 
> I can provide a simple C++ code. But you will need Intel C++ compiler to
> see the problem. Also the current Intel compiler doesn't support RedHat
> 9.

That's good enough - I'd appreciate it if you would post an appropriate
testcase or at least test program.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: RFC: DW_TAG_try_block

[Elena Zannoni]

H. J. Lu writes:
 > On Tue, May 27, 2003 at 10:47:03AM -0400, Elena Zannoni wrote:
 > > H. J. Lu writes:
 > >  > Intel C++ compiler generates DW_TAG_try_block for
 > >  > 
 > >  > 	try
 > >  > 	  {
 > >  > 	  ...
 > >  > 	  }
 > >  > 
 > >  > Gdb can't print any local variables declared inside the try block. This
 > >  > patch seems to work for us. Is that correct?
 > >  > 
 > >  > 
 > >  > H.J.
 > >  > ---
 > >  > 2003-05-26  H.J. Lu <hongjiu.lu@intel.com>
 > >  > 
 > >  > 	* dwarf2read.c (process_die): Handle DW_TAG_try_block.
 > >  > 
 > >  > --- gdb/dwarf2read.c.try	2003-05-26 12:23:22.000000000 -0700
 > >  > +++ gdb/dwarf2read.c	2003-05-26 12:37:08.000000000 -0700
 > >  > @@ -1768,6 +1768,7 @@ process_die (struct die_info *die, struc
 > >  >           of a function and make GDB `next' properly over inlined functions.  */
 > >  >        break;
 > >  >      case DW_TAG_lexical_block:
 > >  > +    case DW_TAG_try_block:
 > >  >        read_lexical_block_scope (die, objfile, cu_header);
 > >  >        break;
 > >  >      case DW_TAG_class_type:
 > > 
 > > 
 > > Seems ok, can you address Daniel's concern about the catch block?
 > > I.e. add that as well? 
 > 
 > http://sources.redhat.com/ml/gdb/2003-05/msg00348.html
 > 

oh, ok then.

 > > How about adding a little test case to the gdb.c++ directory?
 > > 
 > 
 > I can provide a simple C++ code. But you will need Intel C++ compiler to
 > see the problem. Also the current Intel compiler doesn't support RedHat
 > 9.
 > 

Hmm, maybe just post the example code then.

elena


 > 
 > H.J.

Re: RFC: DW_TAG_try_block

[H. J. Lu]

On Tue, May 27, 2003 at 11:12:12AM -0400, Elena Zannoni wrote:
>  > > How about adding a little test case to the gdb.c++ directory?
>  > > 
>  > 
>  > I can provide a simple C++ code. But you will need Intel C++ compiler to
>  > see the problem. Also the current Intel compiler doesn't support RedHat
>  > 9.
>  > 
> 
> Hmm, maybe just post the example code then.
> 
> elena
> 

When it is compiled by ecc 7.1,

# ecc -g main.cc -o main

The current gdb won't see ni nor n when break points are set in try
and catch blocks.


H.J.
-----
#include <iostream>
#include <stdexcept>

using namespace std;

int main(int argc, char* argv[])
{
    try {

        int ni = 2;
        if (ni)
            cout << "scatt: " << ni << endl;

        throw "Good: ";

        exit(EXIT_SUCCESS);
    } catch(runtime_error& e) {
        cerr << "scatt: runtime error occured.\n";
        exit(EXIT_FAILURE);
    }
    catch (const char *p)
    {
        int n;
        n = 3;
        cerr << p << n << endl;
        exit(EXIT_FAILURE);
    }
}

Re: RFC: DW_TAG_try_block

[David Carlton]

On Tue, 27 May 2003 08:13:34 -0700, "H. J. Lu" <hjl@lucon.org> said:

> When it is compiled by ecc 7.1,

> # ecc -g main.cc -o main

> The current gdb won't see ni nor n when break points are set in try
> and catch blocks.

[ Source code deleted, but thanks. ]

I was going to volunteer to add this to the test suite, but it looks
like something like that is already there: am I correct in thinking
that gdb.c++/try-catch.exp fails if you're using ecc as your compiler?
(How do you tell runtest to use a different C compiler, anyways?  Does
"CC=ecc runtest gdb.c++/try-catch.exp" work?)

David Carlton
carlton@math.stanford.edu

Re: RFC: DW_TAG_try_block

[Daniel Jacobowitz]

On Wed, May 28, 2003 at 09:46:48AM -0700, David Carlton wrote:
> On Tue, 27 May 2003 08:13:34 -0700, "H. J. Lu" <hjl@lucon.org> said:
> 
> > When it is compiled by ecc 7.1,
> 
> > # ecc -g main.cc -o main
> 
> > The current gdb won't see ni nor n when break points are set in try
> > and catch blocks.
> 
> [ Source code deleted, but thanks. ]
> 
> I was going to volunteer to add this to the test suite, but it looks
> like something like that is already there: am I correct in thinking
> that gdb.c++/try-catch.exp fails if you're using ecc as your compiler?
> (How do you tell runtest to use a different C compiler, anyways?  Does
> "CC=ecc runtest gdb.c++/try-catch.exp" work?)

No; I have to create a custom board file and use "set board_info
c++compiler ecc" or the equivalent.  Fixing it to recognize
CC_FOR_TARGET is on my list somewhere.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer