Re: Issue 1732 in lilypond: occasional segfault when compiling input/regression/midi/key-initial.ly

[David Kastrup]

Mike Solomon <address@hidden> writes:

> On Aug 23, 2011, at 6:02 PM, address@hidden wrote:
>
>> 
>> Okay, I revise my comment 13. It looks different. With that
>> backtrace (notice this=0x12 in Grob::internal_set_property!) it
>> rather looks like the Engraver_group::acknowledge_grobs has a
>> Grob_info entry in the announce_infos_ list with an uninitialized
>> value (or some wrong value, or something else has gone wrong
>> there... But then, Engraver_group::acknowledge_grobs should have
>> failed already in line 102...).
>> 
>> 
>
> valgrind has a --track-origins=yes flag that allows you to track down
> the origin of uninitialized values . Maybe this'll help?

You noticed
==7941== Memcheck, a memory error detector
==7941== Copyright (C) 2002-2010, and GNU GPL'd, by Julian Seward et al.
==7941== Using Valgrind-3.6.1 and LibVEX; rerun with -h for copyright info
==7941== Command: out/bin/lilypond input/regression/tablature-string-tunings.ly
==7941== Parent PID: 15967
==7941== 
--7941-- 
--7941-- Valgrind options:
--7941--    --suppressions=/usr/lib/valgrind/debian-libc6-dbg.supp
--7941--    --malloc-fill=aa
--7941--    --read-var-info=yes
--7941--    -v
--7941--    --track-origins=yes
--7941--    --log-file=/tmp/lily.log

in the valgrind log file I posted?  The malloc-fill made no difference,
by the way.

-- 
David Kastrup