Bug#610664: minimal recipe to replicate?

2011-01-29 Thread Julian Andres Klode
On Fri, Jan 28, 2011 at 07:37:29PM +, Adam D. Barratt wrote:
 On Mon, 2011-01-24 at 12:06 -0500, Yaroslav Halchenko wrote:
  would it fail with a blank 'int main() {return 0;}'?   I just wonder on 
  which
  program it could be reliably replicated?  Could you please provide a 
  complete
  recipe?
 
 Pedro, would you be able to provide the information which Yaroslav asked for, 
 please?
 
  I have rebuilt
  google-perftools-1.5
  with noopt,nostrip and linked vw (vowpal-wabbit) against profiler -- no
  segfaults...
  
  NB I have not tried in a clean squeeze... just on my 
  squeeze/sid/experimental box
 
 Was that on i386 and amd64?  Having had a look at the package this
 evening, it's README.gz says:
 
I did a similar thing and ran several programs (midori, gnome-about, vlc)
using LD_PRELOAD, simple test programs linked using -lprofiler, and the
complete test suite and could not produce any crash (all on amd64).

I think this bug could be downgraded to 'important', as it works for
others, and could just be an application issue and not a perftools
bug at all (e.g. access via an uninitialized pointer, given that this
points to 0xb5=231 which is very small and not correctly aligned for
an object of that type).
-- 
Julian Andres Klode  - Debian Developer, Ubuntu Member

See http://wiki.debian.org/JulianAndresKlode and http://jak-linux.org/.



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#610664: minimal recipe to replicate?

2011-01-29 Thread Julien Cristau
severity 610664 important
kthxbye

On Sat, Jan 29, 2011 at 15:20:20 +0100, Julian Andres Klode wrote:

 I did a similar thing and ran several programs (midori, gnome-about, vlc)
 using LD_PRELOAD, simple test programs linked using -lprofiler, and the
 complete test suite and could not produce any crash (all on amd64).
 
 I think this bug could be downgraded to 'important', as it works for
 others, and could just be an application issue and not a perftools
 bug at all (e.g. access via an uninitialized pointer, given that this
 points to 0xb5=231 which is very small and not correctly aligned for
 an object of that type).

Downgrading then.

Cheers,
Julien


signature.asc
Description: Digital signature


Bug#610664: minimal recipe to replicate?

2011-01-28 Thread Adam D. Barratt
On Mon, 2011-01-24 at 12:06 -0500, Yaroslav Halchenko wrote:
 would it fail with a blank 'int main() {return 0;}'?   I just wonder on which
 program it could be reliably replicated?  Could you please provide a complete
 recipe?

Pedro, would you be able to provide the information which Yaroslav asked for, 
please?

 I have rebuilt
 google-perftools-1.5
 with noopt,nostrip and linked vw (vowpal-wabbit) against profiler -- no
 segfaults...
 
 NB I have not tried in a clean squeeze... just on my squeeze/sid/experimental 
 box

Was that on i386 and amd64?  Having had a look at the package this
evening, it's README.gz says:

quote
2) On x86-64 64-bit systems, while tcmalloc itself works fine, the
cpu-profiler tool is unreliable: it will sometimes work, but sometimes
cause a segfault.  I'll explain the problem first, and then some
workarounds.

Note that this only affects the cpu-profiler, which is a
google-perftools feature you must turn on manually by setting the
CPUPROFILE environment variable.  If you do not turn on cpu-profiling,
you shouldn't see any crashes due to perftools.
/quote

Regards,

Adam




-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#610664: minimal recipe to replicate?

2011-01-24 Thread Yaroslav Halchenko
Hi Guys,

would it fail with a blank 'int main() {return 0;}'?   I just wonder on which
program it could be reliably replicated?  Could you please provide a complete
recipe?


I have rebuilt
google-perftools-1.5
with noopt,nostrip and linked vw (vowpal-wabbit) against profiler -- no
segfaults...

NB I have not tried in a clean squeeze... just on my squeeze/sid/experimental 
box

according to backtrace -- could it be actually related to boost library so that
illegal access just gets exercised by the profiler, while original
program simply never access that memory... is valgrind silent while running on
the program (without linked profiler)?


-- 
=--=
Keep in touch www.onerussian.com
Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org