Re: Core dump with combination of tie, fermata and some pitches

[Thomas Morley]

2016-09-02 20:12 GMT+02:00 David Kastrup <address@hidden>:
> Timothy Lanfear <address@hidden> writes:
>
>> On 02/09/16 17:31, David Kastrup wrote:
>>> Timothy Lanfear <address@hidden> writes:
>>>
>>>> Bug appears in 2.19.46 and .47. Ubuntu 16.04, 64-bit.
>>>>
>>>> The first line provokes a core dump. The second line is perhaps the
>>>> right syntax for this construction and works fine, but I'm reporting
>>>> in case there is something here that needs investigating.
>>>>
>>>> \version "2.19.46"
>>>>
>>>> {
>>>>    <b' gis'>2\fermata ~ b'2  % core dump
>>>>    <b'~ gis'>2\fermata b'2   % move tie inside chord, OK
>>>>    <b' g'>2\fermata ~ b'2    % gis => g, OK
>>>>    <b' gis'>2 ~ b'2          % delete fermata, OK
>>>>    <b' gis'>2\fermata b'2    % delete tie, OK
>>>> }
>>> Not here.  Which compilations are you using?  I suspect some compiler
>>> bug or overoptimization.
>>>
>>
>> I'm using the pre-built package.
>>
>> address@hidden /usr/local/lilypond-2.19.46/bin/lilypond tmp.ly
>> GNU LilyPond 2.19.46
>> Processing `tmp.ly'
>> Parsing...
>> Interpreting music...
>> Preprocessing graphical objects...Segmentation fault (core dumped)
>
> Huh.  Don't remember the compiler version used currently in Gub.  I'll
> try a native compilation with gcc 6.x 64-bit here and see where this
> gets me.
>
> Man, we really have a lot of infrastructure hickups lately.
>
> --
> David Kastrup



I can confirm the bug with a build from
commit d6e93146b9f25e914373caeced7314237cc53e2f
(Don't remember which optimizing-option I used, though)

gdb says
Program received signal SIGSEGV, Segmentation fault.
0x00000000004ae5ab in Grob::get_offset (a=X_AXIS, this=0x0) at
/home/hermann/lilypond-git/lily/grob.cc:401
401      if (dim_cache_[a].offset_)

complete output from
(gdb) bt
attached.

Cheers,
  Harm

atest-40-bt-log.txt
Description: Text document