[RFC] The -Og debugging experience

[RFC] The -Og debugging experience

[Jiri Olsa]

heya,
using the -Og for DEBUG=1 builds gives me many 'optimized out' stuff

It was introduced in here:
  e8b7ea4356fd perf tools: Improve setting of gcc debug option

- here's backtrace from segfault I was looking at, current code:

(gdb) bt
#0  0x00000000004b490f in parse_events__scan_string (yystr=yystr@entry=0x0, yyscanner=0x1916be0) at util/parse-events-flex.c:3353
#1  0x000000000048fe6f in parse_events__scanner (str=0x0, data=data@entry=0x7fffffffdac0, start_token=start_token@entry=258)
    at util/parse-events.c:1356
#2  0x0000000000491775 in parse_events (evlist=evlist@entry=0x1917c90, str=<optimized out>, err=err@entry=0x0) at util/parse-events.c:1401
#3  0x0000000000468bf6 in __perf_evsel__name_array_test (names=0x872c00 <perf_evsel.sw_names>, nr_names=nr_names@entry=11)
    at tests/evsel-roundtrip-name.c:74
#4  0x0000000000468f13 in test__perf_evsel__roundtrip_name_test (subtest=<optimized out>) at tests/evsel-roundtrip-name.c:106
#5  0x000000000045a2c8 in run_test (test=test@entry=0x86fe50 <generic_tests+400>, subtest=subtest@entry=-1) at tests/builtin-test.c:241
#6  0x000000000045a3a1 in test_and_print (t=t@entry=0x86fe50 <generic_tests+400>, force_skip=force_skip@entry=false,
    subtest=subtest@entry=-1) at tests/builtin-test.c:268
#7  0x000000000045a5cd in __cmd_test (argc=argc@entry=1, argv=argv@entry=0x7fffffffe200, skiplist=0x0) at tests/builtin-test.c:324
#8  0x000000000045a8a9 in cmd_test (argc=1, argv=0x7fffffffe200, prefix=<optimized out>) at tests/builtin-test.c:416
#9  0x00000000004784a9 in run_builtin (p=p@entry=0x871e18 <commands+504>, argc=argc@entry=2, argv=argv@entry=0x7fffffffe200) at perf.c:387
#10 0x00000000004786a4 in handle_internal_command (argc=2, argv=0x7fffffffe200) at perf.c:448
#11 0x0000000000478710 in run_argv (argcp=argcp@entry=0x7fffffffe06c, argv=argv@entry=0x7fffffffe060) at perf.c:492
#12 0x0000000000478981 in main (argc=2, argv=0x7fffffffe200) at perf.c:609


- and same crash with above patch reverted:

(gdb) bt
#0  0x00007ffff573f36a in strlen () from /lib64/libc.so.6
#1  0x00000000004f458b in parse_events__scan_string (yystr=0x0, yyscanner=0x1985be0) at util/parse-events-flex.c:3353
#2  0x00000000004bc752 in parse_events__scanner (str=0x0, data=0x7fffffffda50, start_token=258) at util/parse-events.c:1356
#3  0x00000000004bc8e2 in parse_events (evlist=0x1986c90, str=0x0, err=0x0) at util/parse-events.c:1401
#4  0x00000000004837c4 in __perf_evsel__name_array_test (names=0x8e1140 <perf_evsel.sw_names>, nr_names=11)
    at tests/evsel-roundtrip-name.c:74
#5  0x0000000000483969 in test__perf_evsel__roundtrip_name_test (subtest=-1) at tests/evsel-roundtrip-name.c:106
#6  0x000000000047099b in run_test (test=0x8de030 <generic_tests+400>, subtest=-1) at tests/builtin-test.c:241
#7  0x0000000000470ae3 in test_and_print (t=0x8de030 <generic_tests+400>, force_skip=false, subtest=-1) at tests/builtin-test.c:268
#8  0x0000000000470d71 in __cmd_test (argc=1, argv=0x7fffffffe200, skiplist=0x0) at tests/builtin-test.c:324
#9  0x00000000004711c4 in cmd_test (argc=1, argv=0x7fffffffe200, prefix=0x0) at tests/builtin-test.c:416
#10 0x0000000000498070 in run_builtin (p=0x8dfe58 <commands+504>, argc=2, argv=0x7fffffffe200) at perf.c:387
#11 0x00000000004982de in handle_internal_command (argc=2, argv=0x7fffffffe200) at perf.c:448
#12 0x000000000049842a in run_argv (argcp=0x7fffffffe05c, argv=0x7fffffffe050) at perf.c:492
#13 0x0000000000498786 in main (argc=2, argv=0x7fffffffe200) at perf.c:609


I haven't tracked the history of -Og, maybe it gets better from some gcc version?

If there's no clue, I rather revert this one, because it does
provide the proper debugging experience ;-)

thoughts?
jirka

Re: [RFC] The -Og debugging experience

[Ingo Molnar]

* Jiri Olsa <jolsa@redhat.com> wrote:

> If there's no clue, I rather revert this one, because it does
> provide the proper debugging experience ;-)

Agreed absolutely - I didn't realize that it's broken.

Thanks,

	Ingo

Re: [RFC] The -Og debugging experience

[Martin Liška]

On 12/02/2015 05:48 PM, Jiri Olsa wrote:
> heya,
> using the -Og for DEBUG=1 builds gives me many 'optimized out' stuff
> 
> It was introduced in here:
>   e8b7ea4356fd perf tools: Improve setting of gcc debug option
> 
> - here's backtrace from segfault I was looking at, current code:

Hi.

Can you please provide a test-case which I can use for testing of the issue?
If -Og is problem, I would suggest to use -O0 as a default DEBUG configure option.

Thanks,
Martin

> 
> (gdb) bt
> #0  0x00000000004b490f in parse_events__scan_string (yystr=yystr@entry=0x0, yyscanner=0x1916be0) at util/parse-events-flex.c:3353
> #1  0x000000000048fe6f in parse_events__scanner (str=0x0, data=data@entry=0x7fffffffdac0, start_token=start_token@entry=258)
>     at util/parse-events.c:1356
> #2  0x0000000000491775 in parse_events (evlist=evlist@entry=0x1917c90, str=<optimized out>, err=err@entry=0x0) at util/parse-events.c:1401
> #3  0x0000000000468bf6 in __perf_evsel__name_array_test (names=0x872c00 <perf_evsel.sw_names>, nr_names=nr_names@entry=11)
>     at tests/evsel-roundtrip-name.c:74
> #4  0x0000000000468f13 in test__perf_evsel__roundtrip_name_test (subtest=<optimized out>) at tests/evsel-roundtrip-name.c:106
> #5  0x000000000045a2c8 in run_test (test=test@entry=0x86fe50 <generic_tests+400>, subtest=subtest@entry=-1) at tests/builtin-test.c:241
> #6  0x000000000045a3a1 in test_and_print (t=t@entry=0x86fe50 <generic_tests+400>, force_skip=force_skip@entry=false,
>     subtest=subtest@entry=-1) at tests/builtin-test.c:268
> #7  0x000000000045a5cd in __cmd_test (argc=argc@entry=1, argv=argv@entry=0x7fffffffe200, skiplist=0x0) at tests/builtin-test.c:324
> #8  0x000000000045a8a9 in cmd_test (argc=1, argv=0x7fffffffe200, prefix=<optimized out>) at tests/builtin-test.c:416
> #9  0x00000000004784a9 in run_builtin (p=p@entry=0x871e18 <commands+504>, argc=argc@entry=2, argv=argv@entry=0x7fffffffe200) at perf.c:387
> #10 0x00000000004786a4 in handle_internal_command (argc=2, argv=0x7fffffffe200) at perf.c:448
> #11 0x0000000000478710 in run_argv (argcp=argcp@entry=0x7fffffffe06c, argv=argv@entry=0x7fffffffe060) at perf.c:492
> #12 0x0000000000478981 in main (argc=2, argv=0x7fffffffe200) at perf.c:609
> 
> 
> - and same crash with above patch reverted:
> 
> (gdb) bt
> #0  0x00007ffff573f36a in strlen () from /lib64/libc.so.6
> #1  0x00000000004f458b in parse_events__scan_string (yystr=0x0, yyscanner=0x1985be0) at util/parse-events-flex.c:3353
> #2  0x00000000004bc752 in parse_events__scanner (str=0x0, data=0x7fffffffda50, start_token=258) at util/parse-events.c:1356
> #3  0x00000000004bc8e2 in parse_events (evlist=0x1986c90, str=0x0, err=0x0) at util/parse-events.c:1401
> #4  0x00000000004837c4 in __perf_evsel__name_array_test (names=0x8e1140 <perf_evsel.sw_names>, nr_names=11)
>     at tests/evsel-roundtrip-name.c:74
> #5  0x0000000000483969 in test__perf_evsel__roundtrip_name_test (subtest=-1) at tests/evsel-roundtrip-name.c:106
> #6  0x000000000047099b in run_test (test=0x8de030 <generic_tests+400>, subtest=-1) at tests/builtin-test.c:241
> #7  0x0000000000470ae3 in test_and_print (t=0x8de030 <generic_tests+400>, force_skip=false, subtest=-1) at tests/builtin-test.c:268
> #8  0x0000000000470d71 in __cmd_test (argc=1, argv=0x7fffffffe200, skiplist=0x0) at tests/builtin-test.c:324
> #9  0x00000000004711c4 in cmd_test (argc=1, argv=0x7fffffffe200, prefix=0x0) at tests/builtin-test.c:416
> #10 0x0000000000498070 in run_builtin (p=0x8dfe58 <commands+504>, argc=2, argv=0x7fffffffe200) at perf.c:387
> #11 0x00000000004982de in handle_internal_command (argc=2, argv=0x7fffffffe200) at perf.c:448
> #12 0x000000000049842a in run_argv (argcp=0x7fffffffe05c, argv=0x7fffffffe050) at perf.c:492
> #13 0x0000000000498786 in main (argc=2, argv=0x7fffffffe200) at perf.c:609
> 
> 
> I haven't tracked the history of -Og, maybe it gets better from some gcc version?
> 
> If there's no clue, I rather revert this one, because it does
> provide the proper debugging experience ;-)
> 
> thoughts?
> jirka
> 

Re: [RFC] The -Og debugging experience

[Jiri Olsa]

On Mon, Dec 07, 2015 at 11:48:07AM +0100, Martin Liška wrote:
> On 12/02/2015 05:48 PM, Jiri Olsa wrote:
> > heya,
> > using the -Og for DEBUG=1 builds gives me many 'optimized out' stuff
> > 
> > It was introduced in here:
> >   e8b7ea4356fd perf tools: Improve setting of gcc debug option
> > 
> > - here's backtrace from segfault I was looking at, current code:
> 
> Hi.
> 
> Can you please provide a test-case which I can use for testing of the issue?

just run perf report in gdb and kill it with sigsegv from other terminal
this gives me several instances of <optimized out> right in the backtrace

jirka

Re: [RFC] The -Og debugging experience

[Martin Liška]

On 12/07/2015 02:45 PM, Jiri Olsa wrote:
> On Mon, Dec 07, 2015 at 11:48:07AM +0100, Martin Liška wrote:
>> On 12/02/2015 05:48 PM, Jiri Olsa wrote:
>>> heya,
>>> using the -Og for DEBUG=1 builds gives me many 'optimized out' stuff
>>>
>>> It was introduced in here:
>>>   e8b7ea4356fd perf tools: Improve setting of gcc debug option
>>>
>>> - here's backtrace from segfault I was looking at, current code:
>>
>> Hi.
>>
>> Can you please provide a test-case which I can use for testing of the issue?
> 
> just run perf report in gdb and kill it with sigsegv from other terminal
> this gives me several instances of <optimized out> right in the backtrace
> 
> jirka
> 

Hi.

Unfortunately, running 'perf report' for a medium-size report is very fast a
killing the process from other terminal produces:

[Current thread is 1 (Thread 0x7ffff7fc5740 (LWP 7429))]
(gdb) bt
#0  0x00007ffff632d230 in __write_nocancel () from /lib64/libc.so.6
#1  0x00007ffff62c4dff in _IO_new_file_write () from /lib64/libc.so.6
#2  0x00007ffff62c4403 in new_do_write () from /lib64/libc.so.6
#3  0x00007ffff62c5d09 in __GI__IO_do_write () from /lib64/libc.so.6
#4  0x00007ffff62c5417 in __GI__IO_file_xsputn () from /lib64/libc.so.6
#5  0x00007ffff6299cdb in vfprintf () from /lib64/libc.so.6
#6  0x00007ffff62a03f7 in fprintf () from /lib64/libc.so.6
#7  0x00000000004ef12c in hist_entry__fprintf (he=he@entry=0x1c62b30, size=<optimized out>, size@entry=0, hists=hists@entry=0x1813818, 
    bf=bf@entry=0x1d5cd40 "     0.08%  cc1plus   cc1plus", ' ' <repeats 11 times>, "[.] _Z25number_of_iterations_exitP4loopP8edge_defP15tree_niter_descbb", ' ' <repeats 91 times>..., bfsz=bfsz@entry=479, fp=fp@entry=0x7ffff65f0640 <_IO_2_1_stdout_>)
    at ui/stdio/hist.c:427
#8  0x00000000004ef549 in hists__fprintf (hists=hists@entry=0x1813818, show_header=show_header@entry=true, max_rows=max_rows@entry=0, max_cols=max_cols@entry=0, min_pcnt=0, fp=0x7ffff65f0640 <_IO_2_1_stdout_>) at ui/stdio/hist.c:534
#9  0x000000000042d6a3 in perf_evlist__tty_browse_hists (evlist=0x1812c90, rep=rep@entry=0x7fffffffc6e0, help=help@entry=0x515948 "For a higher level overview, try: perf report --sort comm,dso") at builtin-report.c:370
#10 0x000000000042d7d2 in report__browse_hists (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:455
#11 0x000000000042d992 in __cmd_report (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:571
#12 0x000000000042ec1f in cmd_report (argc=0, argv=0x7fffffffde00, prefix=<optimized out>) at builtin-report.c:957
#13 0x000000000046c496 in run_builtin (p=p@entry=0x7771a0 <commands+192>, argc=argc@entry=1, argv=argv@entry=0x7fffffffde00) at perf.c:387
#14 0x000000000046c693 in handle_internal_command (argc=1, argv=0x7fffffffde00) at perf.c:448
#15 0x000000000046c6fe in run_argv (argcp=argcp@entry=0x7fffffffdc6c, argv=argv@entry=0x7fffffffdc60) at perf.c:492
#16 0x000000000046c94c in main (argc=1, argv=0x7fffffffde00) at perf.c:609

Which is fine.

I've been using GCC 5.2. What version are you using?
I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.

I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.
Thanks,
Martin

Re: [RFC] The -Og debugging experience

[Jiri Olsa]

On Mon, Dec 07, 2015 at 03:04:27PM +0100, Martin Liška wrote:

SNIP

> Hi.
> 
> Unfortunately, running 'perf report' for a medium-size report is very fast a

not if you use TUI ;-)

> killing the process from other terminal produces:
> 
> [Current thread is 1 (Thread 0x7ffff7fc5740 (LWP 7429))]
> (gdb) bt
> #0  0x00007ffff632d230 in __write_nocancel () from /lib64/libc.so.6
> #1  0x00007ffff62c4dff in _IO_new_file_write () from /lib64/libc.so.6
> #2  0x00007ffff62c4403 in new_do_write () from /lib64/libc.so.6
> #3  0x00007ffff62c5d09 in __GI__IO_do_write () from /lib64/libc.so.6
> #4  0x00007ffff62c5417 in __GI__IO_file_xsputn () from /lib64/libc.so.6
> #5  0x00007ffff6299cdb in vfprintf () from /lib64/libc.so.6
> #6  0x00007ffff62a03f7 in fprintf () from /lib64/libc.so.6
> #7  0x00000000004ef12c in hist_entry__fprintf (he=he@entry=0x1c62b30, size=<optimized out>, size@entry=0, hists=hists@entry=0x1813818, 
                                                                              ^^^^^^^^^^^^^^
                                                                              

>     bf=bf@entry=0x1d5cd40 "     0.08%  cc1plus   cc1plus", ' ' <repeats 11 times>, "[.] _Z25number_of_iterations_exitP4loopP8edge_defP15tree_niter_descbb", ' ' <repeats 91 times>..., bfsz=bfsz@entry=479, fp=fp@entry=0x7ffff65f0640 <_IO_2_1_stdout_>)
>     at ui/stdio/hist.c:427
> #8  0x00000000004ef549 in hists__fprintf (hists=hists@entry=0x1813818, show_header=show_header@entry=true, max_rows=max_rows@entry=0, max_cols=max_cols@entry=0, min_pcnt=0, fp=0x7ffff65f0640 <_IO_2_1_stdout_>) at ui/stdio/hist.c:534
> #9  0x000000000042d6a3 in perf_evlist__tty_browse_hists (evlist=0x1812c90, rep=rep@entry=0x7fffffffc6e0, help=help@entry=0x515948 "For a higher level overview, try: perf report --sort comm,dso") at builtin-report.c:370
> #10 0x000000000042d7d2 in report__browse_hists (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:455
> #11 0x000000000042d992 in __cmd_report (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:571
> #12 0x000000000042ec1f in cmd_report (argc=0, argv=0x7fffffffde00, prefix=<optimized out>) at builtin-report.c:957
                                                                             ^^^^^^^^^^^^^^


> #13 0x000000000046c496 in run_builtin (p=p@entry=0x7771a0 <commands+192>, argc=argc@entry=1, argv=argv@entry=0x7fffffffde00) at perf.c:387
> #14 0x000000000046c693 in handle_internal_command (argc=1, argv=0x7fffffffde00) at perf.c:448
> #15 0x000000000046c6fe in run_argv (argcp=argcp@entry=0x7fffffffdc6c, argv=argv@entry=0x7fffffffdc60) at perf.c:492
> #16 0x000000000046c94c in main (argc=1, argv=0x7fffffffde00) at perf.c:609
> 
> Which is fine.

marked 2 instances of 'optimized out' cases above in your output

> 
> I've been using GCC 5.2. What version are you using?

5.1.1

> I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.
> 
> I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.

if you run TUI, you dont need to be fast ;-) make sure you compile with slang devel pkg

thanks.
jirka

Re: [RFC] The -Og debugging experience

[Martin Liška]

On 12/07/2015 03:16 PM, Jiri Olsa wrote:
> On Mon, Dec 07, 2015 at 03:04:27PM +0100, Martin Liška wrote:
> 
> SNIP
> 
>> Hi.
>>
>> Unfortunately, running 'perf report' for a medium-size report is very fast a
> 
> not if you use TUI ;-)
> 
>> killing the process from other terminal produces:
>>
>> [Current thread is 1 (Thread 0x7ffff7fc5740 (LWP 7429))]
>> (gdb) bt
>> #0  0x00007ffff632d230 in __write_nocancel () from /lib64/libc.so.6
>> #1  0x00007ffff62c4dff in _IO_new_file_write () from /lib64/libc.so.6
>> #2  0x00007ffff62c4403 in new_do_write () from /lib64/libc.so.6
>> #3  0x00007ffff62c5d09 in __GI__IO_do_write () from /lib64/libc.so.6
>> #4  0x00007ffff62c5417 in __GI__IO_file_xsputn () from /lib64/libc.so.6
>> #5  0x00007ffff6299cdb in vfprintf () from /lib64/libc.so.6
>> #6  0x00007ffff62a03f7 in fprintf () from /lib64/libc.so.6
>> #7  0x00000000004ef12c in hist_entry__fprintf (he=he@entry=0x1c62b30, size=<optimized out>, size@entry=0, hists=hists@entry=0x1813818, 
>                                                                               ^^^^^^^^^^^^^^
>                                                                               
> 
>>     bf=bf@entry=0x1d5cd40 "     0.08%  cc1plus   cc1plus", ' ' <repeats 11 times>, "[.] _Z25number_of_iterations_exitP4loopP8edge_defP15tree_niter_descbb", ' ' <repeats 91 times>..., bfsz=bfsz@entry=479, fp=fp@entry=0x7ffff65f0640 <_IO_2_1_stdout_>)
>>     at ui/stdio/hist.c:427
>> #8  0x00000000004ef549 in hists__fprintf (hists=hists@entry=0x1813818, show_header=show_header@entry=true, max_rows=max_rows@entry=0, max_cols=max_cols@entry=0, min_pcnt=0, fp=0x7ffff65f0640 <_IO_2_1_stdout_>) at ui/stdio/hist.c:534
>> #9  0x000000000042d6a3 in perf_evlist__tty_browse_hists (evlist=0x1812c90, rep=rep@entry=0x7fffffffc6e0, help=help@entry=0x515948 "For a higher level overview, try: perf report --sort comm,dso") at builtin-report.c:370
>> #10 0x000000000042d7d2 in report__browse_hists (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:455
>> #11 0x000000000042d992 in __cmd_report (rep=rep@entry=0x7fffffffc6e0) at builtin-report.c:571
>> #12 0x000000000042ec1f in cmd_report (argc=0, argv=0x7fffffffde00, prefix=<optimized out>) at builtin-report.c:957
>                                                                              ^^^^^^^^^^^^^^
> 
> 
>> #13 0x000000000046c496 in run_builtin (p=p@entry=0x7771a0 <commands+192>, argc=argc@entry=1, argv=argv@entry=0x7fffffffde00) at perf.c:387
>> #14 0x000000000046c693 in handle_internal_command (argc=1, argv=0x7fffffffde00) at perf.c:448
>> #15 0x000000000046c6fe in run_argv (argcp=argcp@entry=0x7fffffffdc6c, argv=argv@entry=0x7fffffffdc60) at perf.c:492
>> #16 0x000000000046c94c in main (argc=1, argv=0x7fffffffde00) at perf.c:609
>>
>> Which is fine.
> 
> marked 2 instances of 'optimized out' cases above in your output
> 
>>
>> I've been using GCC 5.2. What version are you using?
> 
> 5.1.1
> 
>> I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.
>>
>> I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.
> 
> if you run TUI, you dont need to be fast ;-) make sure you compile with slang devel pkg
> 
> thanks.
> jirka
> 

Hello.

I've just created PR for GCC:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68836

According to discussing with Jakub Jelinek, that's a semi-known issues that's going to be eventually fixed.

Martin

Re: [RFC] The -Og debugging experience

[Jiri Olsa]

On Thu, Dec 10, 2015 at 02:07:35PM +0100, Martin Liška wrote:

SNIP

> > 
> >> I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.
> >>
> >> I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.
> > 
> > if you run TUI, you dont need to be fast ;-) make sure you compile with slang devel pkg
> > 
> > thanks.
> > jirka
> > 
> 
> Hello.
> 
> I've just created PR for GCC:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68836

cool, thanks

jirka

Re: [RFC] The -Og debugging experience

[Arnaldo Carvalho de Melo]

Em Thu, Dec 10, 2015 at 02:24:00PM +0100, Jiri Olsa escreveu:
> On Thu, Dec 10, 2015 at 02:07:35PM +0100, Martin Liška wrote:
> 
> SNIP
> 
> > > 
> > >> I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.
> > >>
> > >> I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.
> > > 
> > > if you run TUI, you dont need to be fast ;-) make sure you compile with slang devel pkg
> > > 
> > > thanks.
> > > jirka
> > > 
> > 
> > Hello.
> > 
> > I've just created PR for GCC:
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68836
> 
> cool, thanks

I'll add this in the revert commit.

- Arnaldo

Re: [RFC] The -Og debugging experience

[Martin Liška]

On 12/10/2015 04:13 PM, Arnaldo Carvalho de Melo wrote:
> Em Thu, Dec 10, 2015 at 02:24:00PM +0100, Jiri Olsa escreveu:
>> On Thu, Dec 10, 2015 at 02:07:35PM +0100, Martin Liška wrote:
>>
>> SNIP
>>
>>>>
>>>>> I've also tried to run './perf test' and terminate the process at random places, but the back trace was OK.
>>>>>
>>>>> I would appreciate if you send me a patch that causes a segfault that is wrongly displayed.
>>>>
>>>> if you run TUI, you dont need to be fast ;-) make sure you compile with slang devel pkg
>>>>
>>>> thanks.
>>>> jirka
>>>>
>>>
>>> Hello.
>>>
>>> I've just created PR for GCC:
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68836
>>
>> cool, thanks
> 
> I'll add this in the revert commit.
> 
> - Arnaldo
> 

Thanks.

Let's wait for the proper fix before re-enabling -Og optimization level.

Martin

[tip:perf/core] Revert "perf tools: Improve setting of gcc debug option"

[tip-bot for Arnaldo Carvalho de Melo]

Commit-ID:  8488335c039ff4917754332763e21c01a81435b4
Gitweb:     http://git.kernel.org/tip/8488335c039ff4917754332763e21c01a81435b4
Author:     Arnaldo Carvalho de Melo <acme@redhat.com>
AuthorDate: Tue, 8 Dec 2015 16:51:24 -0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Thu, 10 Dec 2015 16:09:40 -0300

Revert "perf tools: Improve setting of gcc debug option"

This reverts commit e8b7ea4356fdd3c4de5478f3418eb84f8dce2b61.

Martin created a gcc PR:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68836

Reported-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Martin Liska <mliska@suse.cz>
Cc: Namhyung Kim <namhyung@kernel.org>,
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20151202164827.GA21124@krava.brq.redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/config/Makefile      |  2 --
 tools/perf/config/utilities.mak | 19 -------------------
 2 files changed, 21 deletions(-)

diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile
index 6eb9a95..a552417 100644
--- a/tools/perf/config/Makefile
+++ b/tools/perf/config/Makefile
@@ -135,8 +135,6 @@ endif
 
 ifeq ($(DEBUG),0)
   CFLAGS += -O6
-else
-  CFLAGS += $(call cc-option,-Og,-O0)
 endif
 
 ifdef PARSER_DEBUG
diff --git a/tools/perf/config/utilities.mak b/tools/perf/config/utilities.mak
index 0ebef09..c16ce83 100644
--- a/tools/perf/config/utilities.mak
+++ b/tools/perf/config/utilities.mak
@@ -177,22 +177,3 @@ $(if $($(1)),$(call _ge_attempt,$($(1)),$(1)),$(call _ge_attempt,$(2)))
 endef
 _ge_attempt = $(if $(get-executable),$(get-executable),$(call _gea_err,$(2)))
 _gea_err  = $(if $(1),$(error Please set '$(1)' appropriately))
-
-# try-run
-# Usage: option = $(call try-run, $(CC)...-o "$$TMP",option-ok,otherwise)
-# Exit code chooses option. "$$TMP" is can be used as temporary file and
-# is automatically cleaned up.
-try-run = $(shell set -e;		\
-	TMP="$(TMPOUT).$$$$.tmp";	\
-	TMPO="$(TMPOUT).$$$$.o";	\
-	if ($(1)) >/dev/null 2>&1;	\
-	then echo "$(2)";		\
-	else echo "$(3)";		\
-	fi;				\
-	rm -f "$$TMP" "$$TMPO")
-
-# cc-option
-# Usage: cflags-y += $(call cc-option,-march=winchip-c6,-march=i586)
-
-cc-option = $(call try-run,\
-	$(CC) $(KBUILD_CPPFLAGS) $(KBUILD_CFLAGS) $(1) -c -x c /dev/null -o "$$TMP",$(1),$(2))