Re: Abrupt end of interpreting music (segmentation fault)

[Przemysław Pawełczyk]

2012/2/19 Przemysław Pawełczyk <address@hidden>:
> On Sun, Feb 19, 2012 at 17:20, Jay Anderson <address@hidden> wrote:
>> On Sun, Feb 19, 2012 at 8:32 AM, Phil Holmes <address@hidden> wrote:
>>> I don't normally install LilyPond on my Linux box - I simply run a
>>> self-compiled version.  It's 64-bit Ubuntu and on this, your file compiles
>>> and runs fine.  So I'm assuming (if you know better, please say) that this
>>> is unlikely to be generic 64-bit, but Debian 64-bit specific.  We'll
>>> certainly pick it up as a bug, but it's likely to be hard to fix, bearing in
>>> mind the fact that our standard development environment is Ubuntu.
>>
>> I also run a 64-bit Ubuntu (11.10, Linux 3.0.0-15-generic) box and
>> this example segfaults on an installed 2.15.30, but not on lilypond
>> built from master. So perhaps this bug was somehow fixed? Every once
>> in a while I get segfaults on 64-bit related to the
>> Span_bar_stub_engraver (I tracked it down to the specific commit at
>> one point: 
>> http://lists.gnu.org/archive/html/lilypond-devel/2012-01/msg00201.html).
>> Adding this into the top of the attached example avoids the segfault
>> in 2.15.30:
>>
>> \layout
>> {
>>  \context
>>  {
>>    \PianoStaff
>>    \remove "Span_bar_stub_engraver"
>>  }
>> }
>>
>> In my case I had a dynamic section which I hadn't filled in yet. I
>> haven't looked at this example too much in depth, but it seems similar
>> enough that the same workaround fixed it.
>
> I'm not sure. As it stands out in my second mail
> http://thread.gmane.org/gmane.comp.gnu.lilypond.bugs/33663/focus=33693
> (copy-pasted summary file of LilyPond x64 tests available here:
> http://paste.przemoc.net/lilypond/abrupt-end-of-interpreting-music/)
> it's not something new, it was there in 2.15.23-27, and magically
> reappeared in 2.15.30. It may be some mandelbug that is really hard to
> track down.
>
> More importantly there are many ways to get rid of this segfault, I
> mentioned a few in the first mail. This makes the matter even more
> vague.

I've cloned git repo, checked out 3d8eae0 specifically
(release/2.15.30-1) and built LilyPond. There is no segfault. So it
looks like it's not a bug in the code after all, but some bug in
building process resulting in a binary that can segfault in some rare
cases. If there were changes in building environment or process,
comparing good versions vs bad ones may help a bit, but it surely will
be hard to fix ultimately.

Regards.

-- 
Przemysław 'Przemoc' Pawełczyk
http://przemoc.net/