question, re: gdb.base/label.exp

question, re: gdb.base/label.exp

[Michael Snyder]

Tom, does this test have a dependency on the compiler version?
It's failing on my RHEL5 system with cvs/head gdb.

Re: question, re: gdb.base/label.exp

[Tom Tromey]

>>>>> "Michael" == Michael Snyder <msnyder@vmware.com> writes:

Michael> Tom, does this test have a dependency on the compiler version?
Michael> It's failing on my RHEL5 system with cvs/head gdb.

It might, but I am not certain.  I looked through the gdb-patches
archives but my original mail didn't say anything; nor did the PR.

You can check pretty easily with `readelf -wi' on the `labels'
executable that is made by the test suite.  Search the output for the
`there' DIE.  I get:

 <2><b7>: Abbrev Number: 8 (DW_TAG_label)
    <b8>   DW_AT_name        : (indirect string, offset: 0xc4): there   
    <bc>   DW_AT_decl_file   : 1        
    <bd>   DW_AT_decl_line   : 13       
    <be>   DW_AT_low_pc      : 0x804840c        

I think the important parts are that (1) it exists and (2) it has a
DW_AT_low_pc.

Tom

Re: question, re: gdb.base/label.exp

[Michael Snyder]

Tom Tromey wrote:
>>>>>> "Michael" == Michael Snyder <msnyder@vmware.com> writes:
> 
> Michael> Tom, does this test have a dependency on the compiler version?
> Michael> It's failing on my RHEL5 system with cvs/head gdb.
> 
> It might, but I am not certain.  I looked through the gdb-patches
> archives but my original mail didn't say anything; nor did the PR.
> 
> You can check pretty easily with `readelf -wi' on the `labels'
> executable that is made by the test suite.  Search the output for the
> `there' DIE.  I get:
> 
>  <2><b7>: Abbrev Number: 8 (DW_TAG_label)
>     <b8>   DW_AT_name        : (indirect string, offset: 0xc4): there   
>     <bc>   DW_AT_decl_file   : 1        
>     <bd>   DW_AT_decl_line   : 13       
>     <be>   DW_AT_low_pc      : 0x804840c        
> 
> I think the important parts are that (1) it exists and (2) it has a
> DW_AT_low_pc.

I see this:

  <2><15d>: Abbrev Number: 8 (DW_TAG_label)
      DW_AT_name        : there
      DW_AT_decl_file   : 1
      DW_AT_decl_line   : 13
      DW_AT_low_pc      : 0x4004f1
  <2><16e>: Abbrev Number: 8 (DW_TAG_label)
      DW_AT_name        : here
      DW_AT_decl_file   : 1
      DW_AT_decl_line   : 16
      DW_AT_low_pc      : 0x4004e0

I'm using gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-50)

Re: question, re: gdb.base/label.exp

[Tom Tromey]

Michael> I see this:
[...]

Looks ok to me.

Can you send the gdb.log file?
Maybe there is an answer in there.

Tom

Re: question, re: gdb.base/label.exp

[Jan Kratochvil]

On Thu, 24 Feb 2011 19:21:28 +0100, Michael Snyder wrote:
>  <2><16e>: Abbrev Number: 8 (DW_TAG_label)
>      DW_AT_name        : here
>      DW_AT_decl_file   : 1
>      DW_AT_decl_line   : 16
************************** ^^ here should be 9
>      DW_AT_low_pc      : 0x4004e0
> 
> I'm using gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-50)

symbol_found() uses just SYMBOL_LINE.

The question is if we say this GCC is broken (and XFAIL it or move it to
gdb.dwarf2/) or whether GDB should use SYMBOL_VALUE_ADDRESS.

I am for the latter, DWARF talks for DW_TAG_label about DW_AT_low_pc and not
about DW_AT_decl_line.


Thanks,
Jan

Re: question, re: gdb.base/label.exp

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> The question is if we say this GCC is broken (and XFAIL it or move it to
Jan> gdb.dwarf2/) or whether GDB should use SYMBOL_VALUE_ADDRESS.

Jan> I am for the latter, DWARF talks for DW_TAG_label about
Jan> DW_AT_low_pc and not about DW_AT_decl_line.

Yes, I also think that makes sense.

Tom

Re: question, re: gdb.base/label.exp

[Jan Kratochvil]

On Thu, 24 Feb 2011 21:09:39 +0100, Tom Tromey wrote:
> >>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> 
> Jan> The question is if we say this GCC is broken (and XFAIL it or move it to
> Jan> gdb.dwarf2/) or whether GDB should use SYMBOL_VALUE_ADDRESS.
> 
> Jan> I am for the latter, DWARF talks for DW_TAG_label about
> Jan> DW_AT_low_pc and not about DW_AT_decl_line.
> 
> Yes, I also think that makes sense.

Just it will not work on reload of the executable changed underneath where PCs
change as EXPLICIT_PC needs to be in charge.

But it already does not work now for `run':

echo 'main(){l:;}'|gcc -g -x c -;../gdb -nx ./a.out
[...]
(gdb) start
[...]
(gdb) b l
[...]
(gdb) cont
Continuing.
Program exited with code 070.
(gdb) run
Starting program: .../a.out 
Error in re-setting breakpoint 2: Function "l" not defined.

symtab_and_line->symtab is too general, there should be symtab_and_line->block.


Regards,
Jan

Re: question, re: gdb.base/label.exp

[Edjunior Barbosa Machado]

On 02/25/2011 06:19 AM, Jan Kratochvil wrote:
> On Thu, 24 Feb 2011 21:09:39 +0100, Tom Tromey wrote:
>>>>>>> "Jan" == Jan Kratochvil<jan.kratochvil@redhat.com>  writes:
>>
>> Jan>  The question is if we say this GCC is broken (and XFAIL it or move it to
>> Jan>  gdb.dwarf2/) or whether GDB should use SYMBOL_VALUE_ADDRESS.
>>
>> Jan>  I am for the latter, DWARF talks for DW_TAG_label about
>> Jan>  DW_AT_low_pc and not about DW_AT_decl_line.
>>
>> Yes, I also think that makes sense.
>
> Just it will not work on reload of the executable changed underneath where PCs
> change as EXPLICIT_PC needs to be in charge.
>
> But it already does not work now for `run':
>

just clarifying, this means it should be fixed on gcc anyway?

I also ran into the same problem on RHEL5 and, although the testcase passes ok on RHEL6, DW_AT_decl_line for label 'here' still seems incorrect (16, instead of 9).

Thanks,
-- 
Edjunior

Re: question, re: gdb.base/label.exp

[Tom Tromey]

>>>>> "Edjunior" == Edjunior Barbosa Machado <emachado@linux.vnet.ibm.com> writes:

Edjunior> just clarifying, this means it should be fixed on gcc anyway?

Edjunior> I also ran into the same problem on RHEL5 and, although the
Edjunior> testcase passes ok on RHEL6, DW_AT_decl_line for label 'here'
Edjunior> still seems incorrect (16, instead of 9).

I changed this code in GDB this past week.  Now GDB looks at the label's
DW_AT_low_pc, not DW_AT_decl_line.

I don't know whether there are bugs with this in older versions of GCC.

Tom

Re: question, re: gdb.base/label.exp

[Jan Kratochvil]

On Tue, 29 Mar 2011 16:54:29 +0200, Tom Tromey wrote:
> I changed this code in GDB this past week.  Now GDB looks at the label's
> DW_AT_low_pc, not DW_AT_decl_line.
> 
> I don't know whether there are bugs with this in older versions of GCC.

At least on RHEL-5 gcc-4.1.2-50.el5.x86_64 the testcase has now turned
FAIL->PASS.


Thanks,
Jan