Better fix for "huge backtrace" problem?

[Denys Vlasenko]

https://bugzilla.redhat.com/show_bug.cgi?id=995889

Synopsis:

--- Running report_uReport ---
Generating core_backtrace
Generating backtrace
Backtrace is too big (33561281 bytes), reducing depth to 512
Backtrace is too big (33561281 bytes), reducing depth to 256
Backtrace is too big (33561281 bytes), reducing depth to 128
Backtrace is too big (33561281 bytes), reducing depth to 64
Backtrace is too big (33555685 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 32
Error: Line 15, column 0: "Thread" header expected
('report_uReport' exited with 1)

The currently implemented solution is to remove -ex disassemble"
gdb command"

args[18] = (char*)"-ex";
args[19] = (char*)"disassemble";
args[20] = NULL;
...
+ /* Disable -ex disassemble, output might be huge preventing backtrace generation */
+ args[18] = NULL;
+ args[19] = NULL;
+

But... disasm sometimes does immediately show the problem...

How about just reducing of disassembly range? Instead
of removing the command, how about:

args[19] = (char*)"disassemble $pc, +64";

?

[Jakub Filak]

Richard, you've told me that it would be much complicated to implement :)

https://github.com/abrt/abrt/pull/716#issuecomment-25525532

Denys, is your simple solution working? Have you tested it?

----- Original Message -----
From: "Denys Vlasenko" <dvlasenk-H+wXaHxf7aLQT0dZR+***@public.gmane.org>
To: "Richard Marko" <rmarko-H+wXaHxf7aLQT0dZR+***@public.gmane.org>, crash-catcher-***@public.gmane.org
Sent: Thursday, October 3, 2013 11:41:22 AM
Subject: Better fix for "huge backtrace" problem?

https://bugzilla.redhat.com/show_bug.cgi?id=995889

Synopsis:

--- Running report_uReport ---
Generating core_backtrace
Generating backtrace
Backtrace is too big (33561281 bytes), reducing depth to 512
Backtrace is too big (33561281 bytes), reducing depth to 256
Backtrace is too big (33561281 bytes), reducing depth to 128
Backtrace is too big (33561281 bytes), reducing depth to 64
Backtrace is too big (33555685 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 32
Error: Line 15, column 0: "Thread" header expected
('report_uReport' exited with 1)

The currently implemented solution is to remove -ex disassemble"
gdb command"

args[18] = (char*)"-ex";
args[19] = (char*)"disassemble";
args[20] = NULL;
...
+ /* Disable -ex disassemble, output might be huge preventing backtrace generation */
+ args[18] = NULL;
+ args[19] = NULL;
+

But... disasm sometimes does immediately show the problem...

How about just reducing of disassembly range? Instead
of removing the command, how about:

args[19] = (char*)"disassemble $pc, +64";

?

[Richard Marko]

Yeah, unless you know how to use gdb :)

Post by Jakub Filak
Richard, you've told me that it would be much complicated to implement :)
https://github.com/abrt/abrt/pull/716#issuecomment-25525532
Denys, is your simple solution working? Have you tested it?
----- Original Message -----
Sent: Thursday, October 3, 2013 11:41:22 AM
Subject: Better fix for "huge backtrace" problem?
https://bugzilla.redhat.com/show_bug.cgi?id=995889
--- Running report_uReport ---
Generating core_backtrace
Generating backtrace
Backtrace is too big (33561281 bytes), reducing depth to 512
Backtrace is too big (33561281 bytes), reducing depth to 256
Backtrace is too big (33561281 bytes), reducing depth to 128
Backtrace is too big (33561281 bytes), reducing depth to 64
Backtrace is too big (33555685 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 64
Backtrace is too big (33555461 bytes), reducing depth to 32
Error: Line 15, column 0: "Thread" header expected
('report_uReport' exited with 1)
The currently implemented solution is to remove -ex disassemble"
gdb command"
args[18] = (char*)"-ex";
args[19] = (char*)"disassemble";
args[20] = NULL;
...
+ /* Disable -ex disassemble, output might be huge preventing backtrace generation */
+ args[18] = NULL;
+ args[19] = NULL;
+
But... disasm sometimes does immediately show the problem...
How about just reducing of disassembly range? Instead
args[19] = (char*)"disassemble $pc, +64";
?