Re: Segmentation Fault (Midi News Flash)

[Carl Sorensen]

On 12/7/11 1:48 PM, "Craig" <address@hidden> wrote:

>Hi Mike,
>
>I have found out what it is.  It has to do with the midi output.  If I
>remove the \midi {} from my project there are not segmentation faults.
>So this has to be a memory allocation issue inside the lily machine.

So now how could anybody at lilypond troubleshoot this?

There is no way for us to reproduce it, so we can't investigate it.

If we solved the problem, how would we know?

Will my system behave the same as your system?  If not, is it because of a
different operating system, a different memory size, or some other reason?

In order for us to even begin to look at the problem you're having, we
need access to a copy of the problematic file.  Even if the file is
*huge*, at least we can see if the problem occurs on our machines as well
as on yours.

If it works on mine, and fails on yours, then we'll need to compare
software and OS versions.  You haven't given that information, so we can't
do that.

Is this segmentation fault due to the number of staves, or the number of
notes, or something else?  Can you eliminate some of the non-tuba staves,
and see if you still get the problem?  Can you eliminate some of the
previous measures in all of the voices/staves and still get the problem?

Can you do something like

myMusic = \repeat unfold 100 {a4 a a a}

\score {
  \new StaffGroup <<
     \new Staff { \myMusic }
     \new Staff { \myMusic }
     \new Staff { \myMusic }
  >>
  \layout {}
  \midi {}
}

and change the number of staves and the number of measures (via \repeat
unfold) and see if you get the error when the number of staves or the
number of measures gets too big?

Since including the \midi {} block causes the problem, can you explore
whether the problem is too many midi voices?

Can you create two separate \score blocks in your file, one with a \layout
{} and the other with a \midi {}, to see if you still get the segfault?

If we don¹t have access to your code, it's like my telling you "My car
won't start.  Any idea what's wrong?"

If we *do* have access to your code, and it's very complicated, we are not
very likely to do lots of detailed troubleshooting, but at least we can
test whether the problem is reproducible.  And we *might* be willing to
try one or two things to see if we can make it work.

If you come to the list with very little information and no code
available, and ask us what's wrong, we can't possibly tell you.  We don't
have any idea.

If you come to the list with that same little bit of information, and ask
us for ideas about how you can troubleshoot it, we might have some
suggestions.  If by following those suggestions, you are able to come up
with a simple (but not necessarily *short*) lilypond file that shows the
behavior, then we have a great chance to troubleshoot it.

For an example of how this worked on another bug that showed up only in a
complex score (Valentin's opera), see Issue 1475
<http://code.google.com/p/lilypond/issues/detail?id=1475&can=1&q=unfold&col
spec=ID%20Type%20Status%20Stars%20Owner%20Patch%20Needs%20Summary>

This started out as a crash only on valentin's opera.  Pal then found a
way to use \repeat unfold to make a simple (but long) score that got the
same error.  Then we could start experimenting simply with the code to
find when the bug was introduced.  And it was a straightforward fix for
the person who knew the right code.

So we can't work miracles to fix your problem.  But we can work with you
to help you get to a state where there's a good enough bug report to fix
the problem.

I hope this helps,

Carl