Re: Check for null pointer

[Dan Eble]

On 2011-08-18, at 10:11 , Carl Sorensen wrote:

> On 8/17/11 10:41 PM, "Dan Eble" <address@hidden> wrote:
> 
>> What I have so far is a backtrace:
>>  http://lists.gnu.org/archive/html/lilypond-devel/2011-08/msg00494.html
>> and a large amount of input spread across many files, which is why I chose to
>> review the lilypond source first.
>> 
>> It may also be of interest that I am generating PartCombineMusic with scheme
>> functions other than the stock part combiner.
>> --
> Two questions:
> 
> 1) Is the segfault repeatable?

After extensive testing, the answer is yes, but it's complicated.  It is 
repeatable, but the result doesn't depend only on the input.

When I build, I prefer to use a program called "color" that captures standard 
output and standard error and highlights the errors.  Here's what happens.  In 
the following list, X is a makefile target, and Y is the lilypond command line 
that "make X" runs.

color make X   -> crash
make X         -> OK
time make X    -> OK

color Y        -> OK
Y              -> OK
time Y         -> OK

make X 2>err.txt >out.txt  -> OK

> 2) If so, can you test it on the latest development release?

I tried Graham's experimental 2.15.9 (from last week) and it worked fine at 
first, but then I tried running it via gdb and it crashed with similar but not 
identical backtrace as 2.14.2.  (The caller of kill_mmrest is 
Part_combine_iterator::process instead of Part_combine_iterator::unisono).
-- 
Dan