Re: Re: Re: problem still exists in the latest 2.15 version

[Neil Puttock]

On 24 February 2012 11:44, Colin Hall <address@hidden> wrote:
>
> On Fri, Feb 24, 2012 at 07:37:55PM +0800, hhpmusic wrote:
>> I tried again on my Win7 64 machine, which has Intel I5 processor and 16GB 
>> RAM for midi sequencing. The problem persists. Strange!
>
> Send whatever output you have.

Haipeng's score segfaults on my 64 bit Ubuntu system.  Looking at the
backtrace, I'd say it's the same fault Jay reported here:
http://lists.gnu.org/archive/html/bug-lilypond/2011-12/msg01167.html

(backtrace)

Program received signal SIGSEGV, Segmentation fault.
0x0000000000614886 in Scheme_hash_table::try_retrieve (this=0x0,
    k=0x7ffff342c140, v=0x7fffffff8188) at scm-hash.cc:97
97        SCM handle = scm_hashq_get_handle (hash_tab_, k);
(gdb) bt
#0  0x0000000000614886 in Scheme_hash_table::try_retrieve (this=0x0,
    k=0x7ffff342c140, v=0x7fffffff8188) at scm-hash.cc:97
#1  0x00000000004ae06d in Context::where_defined (this=0x420f240,
    sym=0x7ffff342c140, value=0x7fffffff8188) at context.cc:420
#2  0x00000000004a7af6 in updated_grob_properties (tg=0x420f240,
    sym=0x7ffff342c140) at context-property.cc:264
#3  0x0000000000668cfe in Span_bar_stub_engraver::process_acknowledged (
    this=0x25f1e40) at span-bar-stub-engraver.cc:119
#4  0x00000000006d6b6d in Translator_method_binding::invoke (this=0x25c8ce0)
    at ./include/translator-group.hh:46
#5  0x00000000006d64a9 in Translator_group::precomputed_translator_foreach (
    this=0x35d8650, idx=PROCESS_ACKNOWLEDGED) at translator-group.cc:325
#6  0x00000000004cb61b in Engraver_group::do_announces (this=0x35d8620)
    at engraver-group.cc:174
#7  0x00000000004cb5dc in Engraver_group::do_announces (this=0x3587b10)
    at engraver-group.cc:169
#8  0x0000000000616116 in Score_engraver::one_time_step (this=0x3587b10)
    at score-engraver.cc:152
#9  0x0000000000615cd5 in Score_engraver::one_time_step_callback (
    self=0x3587b10, ev=0x7fffed9331c0) at score-engraver.cc:145
#10 0x000000000053ff2a in Listener::listen (this=0x13a5230, ev=0x7fffed9331c0)
    at listener.cc:44

There's something suspicious going on in the Span_bar_stub_engraver,
and the thing that sticks out immediately is that the map<Grob *,
Context *> axis_groups_ keeps pointers to contexts which might no
longer exist - if a context dies early (e.g., has no notes to carry it
to the end of the score) and gets removed via RemoveContext, Guile's
garbage collection might sweep it away.  Though there's a check to
ensure a context's alive (via get_parent_context ()), by the time
SpanBarStub objects are created, the first context in
affected_contexts is dead.

Cheers,
Neil