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

[Mike Solomon]

On Aug 23, 2011, at 2:38 PM, address@hidden wrote:

> 
> Comment #13 on issue 1732 by address@hidden: occasional segfault when 
> compiling input/regression/midi/key-initial.ly
> http://code.google.com/p/lilypond/issues/detail?id=1732
> 
> Looking through the code, it appears as follows to me:
> -) Grob::internal_set_property calls Grob::is_live (), which crashed in line 
> 312 with the check
>      return scm_is_pair (immutable_property_alist_);
> -) So, it seems like the immutable_property_alist_ member is not initialized.
> -) All Grob constructors initialize that member from the passed argument 
> (either Grob* or SCM props)
> -) So, my conclusion for now is that somewhere a grob (or some child class 
> instance) is created with the Grob (SCM props) constructor, where the passed 
> argument SCM props is some uninitialized value.
> 

To test this, you can use scm_display (immutable_property_alist_, 
scm_current_error_port ()); in the constructor.  You'll get a lot of garbage, 
but it should (in theory) crash once it hits the uninitialized value.

Cheers,
MS