* Issue with debug symbols with garbage collector
@ 2022-04-20 9:15 Nagaraju Mekala
2022-04-20 19:22 ` Michael Eager
0 siblings, 1 reply; 6+ messages in thread
From: Nagaraju Mekala @ 2022-04-20 9:15 UTC (permalink / raw)
To: gcc-help
Hello All,
When I enable the garbage collector for the microblaze compiler compilation
seems to be fine, but the debug symbols are not generated properly.
If we set a breakpoint at 0x0 it is showing some different location, we can
see the same in the objdump as well.
Disassembly of section .vectors.reset:
00000000 <_start>:
*
*
****************************************************************************/
void microblaze_register_handler(XInterruptHandler Handler, void *DataPtr)
{
MB_InterruptVectorTable[0].Handler = Handler;
0: b0008000 imm -32768
4: b8080000 brai 0
for the _start symbols "void microblaze_register_handler(XInterruptHandler
Handler, void *DataPtr)" is coming up.
If I disable the garbage-collector then everything seems fine.
I have used "-ffunction-sections and -fdata-sections" options during
compilation and -gc-sections during linking command.
Please let me know how to tell the garbage collector to not delete
debug_loc information.
Thanks,
Nagaraju
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: Issue with debug symbols with garbage collector
2022-04-20 9:15 Issue with debug symbols with garbage collector Nagaraju Mekala
@ 2022-04-20 19:22 ` Michael Eager
2022-04-21 7:27 ` Nagaraju Mekala
0 siblings, 1 reply; 6+ messages in thread
From: Michael Eager @ 2022-04-20 19:22 UTC (permalink / raw)
To: Nagaraju Mekala, gcc-help
On 4/20/22 02:15, Nagaraju Mekala via Gcc-help wrote:
> Hello All,
>
> When I enable the garbage collector for the microblaze compiler compilation
> seems to be fine, but the debug symbols are not generated properly.
>
> If we set a breakpoint at 0x0 it is showing some different location, we can
> see the same in the objdump as well.
What different location?
>
>
> Disassembly of section .vectors.reset:
>
> 00000000 <_start>:
> *
> *
> ****************************************************************************/
> void microblaze_register_handler(XInterruptHandler Handler, void *DataPtr)
> {
> MB_InterruptVectorTable[0].Handler = Handler;
> 0: b0008000 imm -32768
> 4: b8080000 brai 0
>
> for the _start symbols "void microblaze_register_handler(XInterruptHandler
> Handler, void *DataPtr)" is coming up.
It looks like both _start and microblaze_register_handler are both at 0.
What did you expect?
GDB picks the first symbol which matches the address.
> If I disable the garbage-collector then everything seems fine.
> I have used "-ffunction-sections and -fdata-sections" options during
> compilation and -gc-sections during linking command.
>
> Please let me know how to tell the garbage collector to not delete
> debug_loc information.
Why do you think that the garbage collector is deleting debug info?
--
Michael Eager
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: Issue with debug symbols with garbage collector
2022-04-20 19:22 ` Michael Eager
@ 2022-04-21 7:27 ` Nagaraju Mekala
2022-04-21 18:11 ` Michael Eager
0 siblings, 1 reply; 6+ messages in thread
From: Nagaraju Mekala @ 2022-04-21 7:27 UTC (permalink / raw)
To: Michael Eager; +Cc: gcc-help
[-- Attachment #1: Type: text/plain, Size: 1920 bytes --]
Hi Michael,
Thanks for the reply.
On Thu, Apr 21, 2022 at 12:52 AM Michael Eager <eager@eagercon.com> wrote:
> On 4/20/22 02:15, Nagaraju Mekala via Gcc-help wrote:
> > Hello All,
> >
> > When I enable the garbage collector for the microblaze compiler
> compilation
> > seems to be fine, but the debug symbols are not generated properly.
> >
> > If we set a breakpoint at 0x0 it is showing some different location, we
> can
> > see the same in the objdump as well.
>
> What different location?
>
> I meant instead of pointing to _start, Debug symbols were pointing to
microblaze_register_handler.
> >
> >
> > Disassembly of section .vectors.reset:
> >
> > 00000000 <_start>:
> > *
> > *
> >
> ****************************************************************************/
> > void microblaze_register_handler(XInterruptHandler Handler, void
> *DataPtr)
> > {
> > MB_InterruptVectorTable[0].Handler = Handler;
> > 0: b0008000 imm -32768
> > 4: b8080000 brai 0
> >
> > for the _start symbols "void
> microblaze_register_handler(XInterruptHandler
> > Handler, void *DataPtr)" is coming up.
>
> It looks like both _start and microblaze_register_handler are both at 0.
> What did you expect?
>
> GDB picks the first symbol which matches the address.
>
I have attached the elf, where we cannot find any symbols related to
microblaze_register_handler function but the debug_loc is somehow showing
up at _start address.
> > If I disable the garbage-collector then everything seems fine.
> > I have used "-ffunction-sections and -fdata-sections" options during
> > compilation and -gc-sections during linking command.
> >
> > Please let me know how to tell the garbage collector to not delete
> > debug_loc information.
>
> Why do you think that the garbage collector is deleting debug info?
>
I was just guessing since _start debug symbols were mixed with
microblaze_register_handler.
> --
> Michael Eager
>
[-- Attachment #2: kc705_hello.elf --]
[-- Type: application/octet-stream, Size: 202256 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: Issue with debug symbols with garbage collector
2022-04-21 7:27 ` Nagaraju Mekala
@ 2022-04-21 18:11 ` Michael Eager
2022-04-21 18:28 ` Michael Eager
0 siblings, 1 reply; 6+ messages in thread
From: Michael Eager @ 2022-04-21 18:11 UTC (permalink / raw)
To: Nagaraju Mekala, Michael Eager; +Cc: gcc-help
On 4/21/22 00:27, Nagaraju Mekala via Gcc-help wrote:
>> I meant instead of pointing to _start, Debug symbols were pointing to
> microblaze_register_handler.
Both symbols are at location 0.
> I have attached the elf, where we cannot find any symbols related to
> microblaze_register_handler function but the debug_loc is somehow showing
> up at _start address.
There are two places where symbols are defined. One is in the ELF
symbol table:
$ readelf -s kc705_hello.elf | grep " _start"
277: 0000000000000000 0 FUNC GLOBAL DEFAULT 1 _start
The other in the DWARF debug_info section:
$ readelf -w kc705_hello.elf
<1><383>: Abbrev Number: 15 (DW_TAG_subprogram)
<384> DW_AT_external : 1
<384> DW_AT_name : (indirect string, offset: 0xf783):
microblaze_register_handler
<388> DW_AT_decl_file : 1
<389> DW_AT_decl_line : 86
<38a> DW_AT_decl_column : 6
<38b> DW_AT_prototyped : 1
<38b> DW_AT_low_pc : 0x0 <=== Fun is at 0x0000
<393> DW_AT_high_pc : 0x18
<39b> DW_AT_frame_base : 1 byte block: 9c
(DW_OP_call_frame_cfa)
<39d> DW_AT_call_all_calls: 1
<39d> DW_AT_sibling : <0x3bc>
GDB uses the DWARF information before falling back to using the
ELF symbol table.
--
Michael Eager
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: Issue with debug symbols with garbage collector
2022-04-21 18:11 ` Michael Eager
@ 2022-04-21 18:28 ` Michael Eager
2022-04-22 6:29 ` Nagaraju Mekala
0 siblings, 1 reply; 6+ messages in thread
From: Michael Eager @ 2022-04-21 18:28 UTC (permalink / raw)
To: Michael Eager, Nagaraju Mekala; +Cc: gcc-help
On 4/21/22 11:11, Michael Eager wrote:
> On 4/21/22 00:27, Nagaraju Mekala via Gcc-help wrote:
>>> I meant instead of pointing to _start, Debug symbols were pointing to
>> microblaze_register_handler.
>
> Both symbols are at location 0.
>
>> I have attached the elf, where we cannot find any symbols related to
>> microblaze_register_handler function but the debug_loc is somehow showing
>> up at _start address.
>
> There are two places where symbols are defined. One is in the ELF
> symbol table:
>
> $ readelf -s kc705_hello.elf | grep " _start"
> 277: 0000000000000000 0 FUNC GLOBAL DEFAULT 1 _start
There are a number of symbols with the prefix microblaze_ which appear
in the ELF symbol table. I can't tell from the very limited information
you provide why those are included but microblaze_register_handler is
not. This probably has to do with how these functions are written.
--
Michael Eager
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Issue with debug symbols with garbage collector
2022-04-21 18:28 ` Michael Eager
@ 2022-04-22 6:29 ` Nagaraju Mekala
0 siblings, 0 replies; 6+ messages in thread
From: Nagaraju Mekala @ 2022-04-22 6:29 UTC (permalink / raw)
To: Michael Eager; +Cc: gcc-help
On Thu, Apr 21, 2022 at 11:58 PM Michael Eager <eager@eagercon.com> wrote:
> On 4/21/22 11:11, Michael Eager wrote:
> > On 4/21/22 00:27, Nagaraju Mekala via Gcc-help wrote:
> >>> I meant instead of pointing to _start, Debug symbols were pointing to
> >> microblaze_register_handler.
> >
> > Both symbols are at location 0.
> >
> >> I have attached the elf, where we cannot find any symbols related to
> >> microblaze_register_handler function but the debug_loc is somehow
> showing
> >> up at _start address.
> >
> > There are two places where symbols are defined. One is in the ELF
> > symbol table:
> >
> > $ readelf -s kc705_hello.elf | grep " _start"
> > 277: 0000000000000000 0 FUNC GLOBAL DEFAULT 1 _start
>
> There are a number of symbols with the prefix microblaze_ which appear
> in the ELF symbol table. I can't tell from the very limited information
> you provide why those are included but microblaze_register_handler is
> not. This probably has to do with how these functions are written.
>
Thanks for the information.
I will check why microblaze_register_handler is not part of symbol table.
-Nagaraju
>
> --
> Michael Eager
>
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2022-04-22 6:29 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-04-20 9:15 Issue with debug symbols with garbage collector Nagaraju Mekala
2022-04-20 19:22 ` Michael Eager
2022-04-21 7:27 ` Nagaraju Mekala
2022-04-21 18:11 ` Michael Eager
2022-04-21 18:28 ` Michael Eager
2022-04-22 6:29 ` Nagaraju Mekala
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).