[PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Martin Galvan]

This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.

This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.

I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.

gdb/
2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>

	* dwarf2read.c (die_needs_namespace): Return 1 for
	DW_TAG_inlined_subroutine.

---
 gdb/dwarf2read.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
index f6b0c01..2bf3513 100644
--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
     case DW_TAG_enumeration_type:
     case DW_TAG_enumerator:
     case DW_TAG_subprogram:
+    case DW_TAG_inlined_subroutine:
     case DW_TAG_member:
     case DW_TAG_imported_declaration:
       return 1;
--
2.3.5

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Doug Evans]

On Thu, Apr 23, 2015 at 8:24 AM, Martin Galvan
<martin.galvan@tallertechnologies.com> wrote:
> This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.
>
> This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.
>
> I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.
>
> gdb/
> 2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>
>
>         * dwarf2read.c (die_needs_namespace): Return 1 for
>         DW_TAG_inlined_subroutine.
>
> ---
>  gdb/dwarf2read.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
> index f6b0c01..2bf3513 100644
> --- a/gdb/dwarf2read.c
> +++ b/gdb/dwarf2read.c
> @@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
>      case DW_TAG_enumeration_type:
>      case DW_TAG_enumerator:
>      case DW_TAG_subprogram:
> +    case DW_TAG_inlined_subroutine:
>      case DW_TAG_member:
>      case DW_TAG_imported_declaration:
>        return 1;

LGTM.
I'll check it in.

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Martin Galvan]

On Thu, Apr 23, 2015 at 12:52 PM, Doug Evans <dje@google.com> wrote:
> LGTM.
> I'll check it in.

Thanks a lot!

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Martin Galvan]

Sorry to bother but I didn't see this committed in the gdb repository.
I don't have write access, so could you please commit this for me?

Thanks a lot!

On Thu, Apr 23, 2015 at 12:52 PM, Doug Evans <dje@google.com> wrote:
> On Thu, Apr 23, 2015 at 8:24 AM, Martin Galvan
> <martin.galvan@tallertechnologies.com> wrote:
>> This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.
>>
>> This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.
>>
>> I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.
>>
>> gdb/
>> 2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>
>>
>>         * dwarf2read.c (die_needs_namespace): Return 1 for
>>         DW_TAG_inlined_subroutine.
>>
>> ---
>>  gdb/dwarf2read.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
>> index f6b0c01..2bf3513 100644
>> --- a/gdb/dwarf2read.c
>> +++ b/gdb/dwarf2read.c
>> @@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
>>      case DW_TAG_enumeration_type:
>>      case DW_TAG_enumerator:
>>      case DW_TAG_subprogram:
>> +    case DW_TAG_inlined_subroutine:
>>      case DW_TAG_member:
>>      case DW_TAG_imported_declaration:
>>        return 1;
>
> LGTM.
> I'll check it in.

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Doug Evans]

On Thu, Apr 23, 2015 at 8:52 AM, Doug Evans <dje@google.com> wrote:
> On Thu, Apr 23, 2015 at 8:24 AM, Martin Galvan
> <martin.galvan@tallertechnologies.com> wrote:
>> This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.
>>
>> This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.
>>
>> I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.
>>
>> gdb/
>> 2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>
>>
>>         * dwarf2read.c (die_needs_namespace): Return 1 for
>>         DW_TAG_inlined_subroutine.
>>
>> ---
>>  gdb/dwarf2read.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
>> index f6b0c01..2bf3513 100644
>> --- a/gdb/dwarf2read.c
>> +++ b/gdb/dwarf2read.c
>> @@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
>>      case DW_TAG_enumeration_type:
>>      case DW_TAG_enumerator:
>>      case DW_TAG_subprogram:
>> +    case DW_TAG_inlined_subroutine:
>>      case DW_TAG_member:
>>      case DW_TAG_imported_declaration:
>>        return 1;
>
> LGTM.
> I'll check it in.

Committed.

I forgot the extra step of setting the author correctly, but it's just
one line so I'm leaving it.

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Martin Galvan]

On Wed, May 13, 2015 at 5:58 PM, Doug Evans <dje@google.com> wrote:
> On Thu, Apr 23, 2015 at 8:52 AM, Doug Evans <dje@google.com> wrote:
>> On Thu, Apr 23, 2015 at 8:24 AM, Martin Galvan
>> <martin.galvan@tallertechnologies.com> wrote:
>>> This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.
>>>
>>> This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.
>>>
>>> I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.
>>>
>>> gdb/
>>> 2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>
>>>
>>>         * dwarf2read.c (die_needs_namespace): Return 1 for
>>>         DW_TAG_inlined_subroutine.
>>>
>>> ---
>>>  gdb/dwarf2read.c | 1 +
>>>  1 file changed, 1 insertion(+)
>>>
>>> diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
>>> index f6b0c01..2bf3513 100644
>>> --- a/gdb/dwarf2read.c
>>> +++ b/gdb/dwarf2read.c
>>> @@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
>>>      case DW_TAG_enumeration_type:
>>>      case DW_TAG_enumerator:
>>>      case DW_TAG_subprogram:
>>> +    case DW_TAG_inlined_subroutine:
>>>      case DW_TAG_member:
>>>      case DW_TAG_imported_declaration:
>>>        return 1;
>>
>> LGTM.
>> I'll check it in.
>
> Committed.
>
> I forgot the extra step of setting the author correctly, but it's just
> one line so I'm leaving it.

Well, considering that it solves a quite important issue and it took
me a really long time to find it, fix it, and then get it commited,
it'd be nice of you if you could fix the author :)

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Doug Evans]

On Wed, May 13, 2015 at 2:04 PM, Martin Galvan
<martin.galvan@tallertechnologies.com> wrote:
> On Wed, May 13, 2015 at 5:58 PM, Doug Evans <dje@google.com> wrote:
>> On Thu, Apr 23, 2015 at 8:52 AM, Doug Evans <dje@google.com> wrote:
>>> On Thu, Apr 23, 2015 at 8:24 AM, Martin Galvan
>>> <martin.galvan@tallertechnologies.com> wrote:
>>>> This bug was being caused by die_needs_namespace returning 0 for a DIE whose tag was DW_TAG_inlined_subroutine. This meant that dwarf2_physname would simply return the DIE's name attribute (which in our case would be "method"). Therefore, when new_symbol_full called SYMBOL_SET_NAMES, the linkagename argument wasn't the demangled name as it should have.
>>>>
>>>> This patch adds a case which would return 1 for DW_TAG_inlined_subroutine in die_needs_namespace. It's tested both for classes and namespaces.
>>>>
>>>> I have a company-wide copyright assignment. I don't have commit access, though, so it would be great if anyone could commit this for me.
>>>>
>>>> gdb/
>>>> 2015-04-23  Martin Galvan  <martin.galvan@tallertechnologies.com>
>>>>
>>>>         * dwarf2read.c (die_needs_namespace): Return 1 for
>>>>         DW_TAG_inlined_subroutine.
>>>>
>>>> ---
>>>>  gdb/dwarf2read.c | 1 +
>>>>  1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
>>>> index f6b0c01..2bf3513 100644
>>>> --- a/gdb/dwarf2read.c
>>>> +++ b/gdb/dwarf2read.c
>>>> @@ -8357,6 +8357,7 @@ die_needs_namespace (struct die_info *die, struct dwarf2_cu *cu)
>>>>      case DW_TAG_enumeration_type:
>>>>      case DW_TAG_enumerator:
>>>>      case DW_TAG_subprogram:
>>>> +    case DW_TAG_inlined_subroutine:
>>>>      case DW_TAG_member:
>>>>      case DW_TAG_imported_declaration:
>>>>        return 1;
>>>
>>> LGTM.
>>> I'll check it in.
>>
>> Committed.
>>
>> I forgot the extra step of setting the author correctly, but it's just
>> one line so I'm leaving it.
>
> Well, considering that it solves a quite important issue and it took
> me a really long time to find it, fix it, and then get it commited,
> it'd be nice of you if you could fix the author :)

Done.

Time to get yourself commit access.

Re: [PING][PATCH] Fix PR gdb/17720 (Function names appear without namespace/class prefixes in backtrace for optimized code)

[Martin Galvan]

On Wed, May 13, 2015 at 6:10 PM, Doug Evans <dje@google.com> wrote:
> On Wed, May 13, 2015 at 2:04 PM, Martin Galvan
> <martin.galvan@tallertechnologies.com> wrote:
>> On Wed, May 13, 2015 at 5:58 PM, Doug Evans <dje@google.com> wrote:
>>>
>>> Committed.
>>>
>>> I forgot the extra step of setting the author correctly, but it's just
>>> one line so I'm leaving it.
>>
>> Well, considering that it solves a quite important issue and it took
>> me a really long time to find it, fix it, and then get it commited,
>> it'd be nice of you if you could fix the author :)
>
> Done.
>
> Time to get yourself commit access.

Thanks a lot!

-- 

Martin Galvan

Software Engineer

Taller Technologies Argentina

San Lorenzo 47, 3rd Floor, Office 5

Córdoba, Argentina

Phone: 54 351 4217888 / +54 351 4218211