Re: [Denemo-devel] Denemo crash

[Richard Shann]

On Sun, 2017-02-05 at 22:08 +0100, Andreas Schneider wrote:
> Am 03.02.2017 um 18:26 schrieb Richard Shann:
> > On Thu, 2017-02-02 at 23:33 +0100, Andreas Schneider wrote:
> >>> Also, if you delete the lyrics from the score I expect the problem
> >> may
> >>> go - it was the snapshotting of the lyrics that had serious bugs.
> >>
> >> When deleting the last lyrics verse, Denemo crashes, even after
> >> executing (d-IncreaseGuard). See attached log file.
> > 
> > This is quite surprising! There was such a bug, a regression from 2.0.14
> > which I fixed on Wed, 1 Feb 2017 14:23:52 +0000
> > 
> > If you set a breakpoint in lyric.c at line 689
> > 
> > b lyric.c:689
> > 
> > the it should stop on that line when you delete the last verse...
> > 
> > and then
> > 
> >  p staff->current_verse_view
> > 
> > should show an empty GList* for the current_verse_view.
> 
> Again, Denemo seems to crash before it reaches the breakpoint, see
> attached log.
Ah, the problem here is that the break point was never set - in your
terminal log you have:

[...]
(gdb) b lyric.c:689
Make breakpoint pending on future shared library load? (y or [n]) 
(gdb) c


The meaning of the question is that gdb has not got any idea where
lyric.c:689 is, and is hoping it is perhaps in some yet-to-be-loaded
library. We *know* it is in the actual program.

I guess you have to have the flag set for debug symbols for the
breakpoint to get set (IIRC the flag is -B for gcc).

... and reading a later email, you turned on debug symbols and the
program didn't crash. This *really* shouldn't happen - the actual
executing code is supposed to be completely unchanged with or without
debug information included - it's like getting a car with or without a
user's manual, the car is supposed to be identical. This would be a gcc
bug, but I would check that you didn't do something else at the same
time as turning on debug symbols (e.g. change -O2 )

Richard