Re: [Rosegarden-devel] trunk crash

[D. M. M. <mic...@ro...>]

On Tuesday 27 November 2007, Pedro Lopez-Cabanillas wrote:

> No. I think that it is caused by NotationGroup::calculateBeam() in
> NotationGroup.cpp:356

The crash is just in the third track/segment.  In bar 11.

Cut compo down to just this bar.

Delete notes at F#3.  Crash.

Delete notes at F#4.  Crash.

Delete note at D4.    Crash.

Delete note at A3.    Crash.

Delete note at G#3.   Crash.

Delete note at G#4.   Crash.

Delete note at A4.    Crash.

Delete note at D5.    Crash.

Delete *all* notes from isolated bar 11.  CRASH.

That's a surprise.  Unless...  Well, if I extend the segment, notes from 
beyond the end do show up (see recent discussion about how this is 
stupid/wrong/confusing) but I'm not sure if that's the answer.

Matrix of segment from bar 11, select all, copy, new segment from scratch, 
matrix, paste.

Opens fine in notation editor.

Ain't that some shit?  The first isolated bar that caused the crash only 
causes the crash when it's a subset of the original segment, with all of the 
right side of the segment contained in its magical hidden area to the right.  
Copy the contents of this one bar out to a new virgin segment, and no crash.

So where's the crash now?

Start over.  Open doc with just segment from bar 11/track2 (plus extra hidden 
tail).  Event list.

Delete all events except four rests.  No crash.

Delete just the final indication.     No crash.

Delete everything except rests (can't be deleted) and the final indication.  
CRASH.

Repeat.  Crash repeated.

Just the final indication (a slur with nothing to attach to in "group5") 
causes a definite crash.

Repeat just deleting this final indication.  No crash.

I'm not sure all of this actually added up to finding the real trigger.  There 
seems to be more than one bug at play here, and I am guessing that I am 
looking at a bug associated with the extra hidden tail stuff, rather than the 
most significant bug.

But maybe.  Reloading the original.  Delete segments from tracks 1 3, open 2 
in event editor and delete the last indication from bar 11.

There is no slur group 5 in bar 11.  It's shown as bar 12.  Delete, and crash 
still occurs.

Back up.  Glazunov.  Track/segment 2.  Event list.  Uncheck everything except 
indication, leaving indications visible.  Delete all of them.  Open in 
notation.

No crash.

So it appears definitely indication-related (or at least one crash in this 
file is.)

Start over.  Delete segment in track 1/3, save as /tmp/crash3.rg.

Open segment in track 2 in event list, show only indications, and:

Delete just the slur from group 5.  Crash.

Delete *all* slurs.                 Crash.

Delete all crecendos.               No crash.

Delete everything but the cresc.    No crash.

This sure is an elusive fucking bug!

Delete everything except notes.     No crash.

I'm bored now.  I'm either looking at multiple crashes for multiple reasons, 
or...  I don't know.  I can't figure it out.  It doesn't seem to be as simple 
as a stem direction thing though.

Does it?

Manually select/delete all events except in bar 11.

Crash.

Just big empty segment with events still in bar 11.  Select everything 
selectable from matrix, delete.  Crash.

Humbug.  I quit.

-- 
D. Michael McIntyre