Re: [Denemo-devel] Denemo crash

[Andreas Schneider]

Am 06.02.2017 um 08:45 schrieb Andreas Schneider:
> Am 05.02.2017 um 22:02 schrieb Andreas Schneider:
>> Am 03.02.2017 um 17:47 schrieb Richard Shann:
>>> On Thu, 2017-02-02 at 23:33 +0100, Andreas Schneider wrote:
>>>> Am 01.02.2017 um 22:48 schrieb Richard Shann:
>>>>> yes, but it could be a bug in Denemo that happens not to be
>>>> triggered. 
>>>>> Can you try executing
>>>>> (d-IncreaseGuard) in the Scheme window or CLI and then seeing if it
>>>>> crashes? d- IncreaseGaurd will turn off the undo, and this will
>>>> indicate
>>>>> that the problem is with the Snapshot done before adding the staff.
>>>>
>>>> After executing (d-IncreaseGuard), the Denemo does not crash any more
>>>> when inserting a staff.
>>>
>>> Well, I've gone over the snapshotting code and done some cleaning and
>>> re-coding, commenting and protected against a condition that shouldn't
>>> arise. This is now in git.
>>> If, with this new version inserting/deleting a staff from that score (I
>>> guess it is any score with lyrics) is still crashing it is going to be
>>> quite tricky to debug.
>>> If you *do* have the current git source code crashing under gdb you
>>> could create a breakpoint at line 1176 of select.c which is where the
>>> snapshot code ends (you do a Ctrl-C to suspend the program, issue
>>>
>>> b select.c:1176
>>>
>>> to insert the break and then the command
>>>
>>> c
>>>
>>> to continue the program).
>>>
>>> Then it should stop on that line when you try to add/delete a staff.
>>>
>>> If it does stop then issuing 
>>>
>>> call call_out_to_guile ("(d-Save \"/home/aschneider/JUNKNAME.denemo\")")
>>>
>>> at the gdb prompt will save the score under the name JUNKNAME.denemo and
>>> this may the hold a clue.
>>
>> Hmm, it seems not to reach that point. I did as you recommended, yet
>> after inserting a new staff Denemo crashes before it reaches the
>> breakpoint. Or did you mean I should do that after executing
>> (d-IncreaseGuard)?
> 
> Oops, I forgot that the package building script automatically strips
> binaries, so the executable did not contain debug informations. After
> building by hand executing the local denemo binary, I was able to save
> the file as recommended. I send it to you in a private mail.

I have now seen that the version with debug informations does not crash.
So it might have something to do with the optimisation?

Andreas